Australian Pork Ltd v Director of Animal & Plant Quarantine

REASONS FOR JUDGMENT WILCOX J: 1 This proceeding concerns two decisions made in relation to the importation of uncooked pig meat into Australia. At issue is the manner in which the decision-makers dealt with the need to protect Australian pig farmers from the risk of contamination of their herds by a disease known as post-weaning multisystemic wasting syndrome ('PMWS'). The proceeding 2 The first applicant is Australian Pork Limited ('APL'), a company formed by Australian pig farmers to represent their interests, including in their relations with governments. The second applicant, Windridge Farms Pty Ltd, owns and operates a number of pig farms near Young, New South Wales. 3 The first respondent is the Director of Animal and Plant Quarantine ('the Director'). On or about 10 May 2004, the then Director, Michael Taylor, decided that future importation of pig meat into Australia 'will be subject to the Quarantine Act 1908 [(Cth) ('the Act')] and the application of measures as specified in the section on Quarantine Requirements in the Final Import Risk Analysis (IRA) Report for Pig Meat (February 2004)'. I will call the report 'the IRAR'. Mr Taylor stated: 'This policy is to be taken into account by decision makers in accordance with the [Act] and Quarantine Proclamation 1998 [(Cth) 'the Proclamation'] as amended'. Mr Taylor's decision ('the IRA Decision') is the first decision under challenge in this proceeding. 4 No issue is taken with that aspect of the IRA Decision which concerns the applicability of the Act and Proclamation to the importation of pig meat. It is the application of the control measures recommended in the IRAR to permit decisions with which the applicants take issue. 5 On or about 30 July 2004, a delegate of the Director granted to Fayman International Pty Ltd ('Fayman') a permit to import uncooked pig meat into Australia from the United States of America between 30 July 2004 and 30 July 2006. The permit was granted subject to conditions that reflected Mr Taylor's decision to adopt the control measures recommended in the IRAR. It is common ground that, in deciding to grant the permit, the delegate applied the IRA Decision. The delegate's decision ('the Permit Decision') is the second decision challenged in this proceeding. 6 Fayman is the second respondent to this proceeding. However, the company filed a submitting appearance and took no part in the hearing of the proceeding. Having regard to that fact, it is convenient to refer to the first respondent, in these reasons, simply as 'the respondent'. 7 The applicants' challenge to the validity of the Permit Decision does not depend upon any matter lying outside the acceptability and adequacy of the IRAR and/or the validity of the IRA Decision. 8 In their amended application, the applicants rely on two bases of the Court's jurisdiction: the Administrative Decisions (Judicial Review) Act 1977 (Cth) ('the ADJR Act') and s 39B of the Judiciary Act 1903 (Cth). The validity of both decisions is challenged on four grounds: (i) error of law; (ii) failure to take into account a relevant consideration; (iii) unreasonableness; and (iv) no evidence. All the grounds are based upon alleged deficiencies in the IRAR. 9 No issue has been raised concerning the applicants' standing to maintain the proceeding. PMWS 10 Professor Roger Morris, who gave expert evidence on behalf of the applicants, described PMWS. Professor Morris is Professor of Animal Health at Massey University in New Zealand. He is also Director of the Massey University EpiCentre, a unit which provides research and consultancy services worldwide in animal health and disease control and offers postgraduate training in epidemiology and related fields such as animal health economics and pig health. 11 Professor Morris obtained degrees of Bachelor of Veterinary Science with Honours from the University of Sydney and Master of Veterinary Science from the University of Melbourne. He spent 11 years at Melbourne University teaching clinical medicine, preventive medicine and epidemiology. During this time, he also undertook clinical veterinary work with cattle, sheep, pigs and horses. 12 In 1977, Professor Morris obtained a Doctor of Philosophy degree from the University of Reading, England, for work on the analysis and computer modelling of disease control strategies and integration of economic evaluation into analytical procedures for disease problems. 13 From 1976 to 1981, Professor Morris was Assistant Director of the Australian Bureau of Animal Health in what is now the Commonwealth Department of Agriculture, Fisheries and Forestry. At various times, he acted as Chief Veterinary Officer of Australia. 14 Professor Morris then spent five years teaching in the United States of America before taking up his present position in 1986. 15 Professor Morris has received numerous awards including Fellowships of epidemiological colleges in several countries. He is well qualified to describe the disease that lies at the heart of this proceeding. 16 In his report, Professor Morris offered this description of PMWS: 'Post-weaning multisystemic wasting syndrome or PMWS is an important emerging disease of pigs. The name was coined in 1996 to describe a wasting syndrome of weaner pigs first identified in 1991 in high health status Canadian herds. Since then the disease has been recognized progressively in the USA, the UK, most of continental Europe, Asia and most recently New Zealand. In affected areas, the number of herds diagnosed with the disease has also increased during this time. Australia stands out among significant pig-raising countries in that it remains free when virtually all others have become infected over less than a decade. As its name suggests, PMWS mainly occurs in weaner pigs (6-12 weeks of age) and is a disease that affects several organ systems. Affected pigs are unthrifty, dyspnoeic (have difficulty breathing), and have enlarged lymph nodes, although this last feature is variable. Less frequently, they have diarrhoea, gastric ulcers and icterus (jaundice). At autopsy the lesions in pigs with PMWS are quite variable, although abnormalities in the lungs and enlargement of some or all of the inguinal, mesenteric, bronchial and mediastinal lymph nodes are consistent findings. Lung lesions vary from failure to collapse and increased firmness to extensive to diffuse red to pale tan mottling with areas of consolidation in anterior ventral areas. Non-collapsing lungs at autopsy represent one of the strong indicators of the disease. Histologically there is multifocal lymphohistiocytic to disseminated granulomatous interstitial pneumonia, often with giant cells. In affected lymph nodes, depletion of lymphocytes together with histiocytic inflammatory infiltration is consistently described. The most characteristic feature of PMWS, although not always present, is the occurrence of intensely basophilic inclusion bodies in many of the affected tissues that are rich in porcine circovirus (PCV) antigen and from which PCV can invariably be isolated.' 17 There are two forms of PCV. The form that received attention in this case is known as PCV2. Professor Morris said this form of the virus 'is ubiquitous in pig herds in all parts of the world examined, and it is rare to find a herd free of the virus'. 18 Professor Morris said PMWS: 'behaves as a propagating epidemic both between herds and within herds. It has moved from country to country in the last decade, in some cases with an identifiable method of entry. At first when it reaches a country only a few herds are affected, but then it starts to spread more rapidly and becomes very difficult to control, especially in pig-dense areas where there are various forms of interchange between herds. High biosecurity is protective for herds, but does not eliminate the possibility of infection. Movement of recently weaned pigs on to a farm is the single most important method of transfer of infection, but feed or equipment contaminated with the agent can probably also spread it to some degree. There are indications that the agent may be capable of windborne transmission between farms over short distances.' 19 After detailing experience with PMWS in Great Britain and Denmark, Professor Morris said: 'Within herds, in some cases the event starts with a period of reproductive problems in sows, but after several weeks disease starts to appear in weaner pigs in the form of wasting and breathing difficulty in a number of pigs, leading to death in about 10 to 14 days. As some pigs die, other previously healthy pigs become affected and also die. Deaths stop when surviving pigs pass out of the susceptible age range. A very characteristic feature of the disease is that antibiotics and other forms of medication have no worthwhile impact on the course of the disease, indicating that the cause is likely to be a virus. Thus all of the evidence points to this disease being caused by a novel virus, but all efforts so far to isolate a new virus have been unsuccessful.' 20 Turning to the impact of PMWS on herds, Professor Morris said: 'Morbidity and mortality are most severe in newly affected herds where 4 to 20% of pigs are usually affected (range 1-60%); of which 70-80% usually die (range 30-100%). After a period which may last from several months to a few years, the disease situation settles down and losses are lower. However herd mortality and growth performance rarely return to pre-disease levels. Herds which trade extensively in pigs may be affected continuously because there are always susceptible pigs entering the farm. The disease is sufficiently severe to force a significant number of affected producers out of business, and the costs of control are substantial in herds which decide to live with the disease. In New Zealand, affected herds have now been carefully assessed for mortality rate in pigs between 4 and 12 weeks of age. Typical mortality rates in unaffected herds are 0.5 to 3%, with some herds kept under adverse environmental conditions higher. Mortality rates in affected herds varied between 10 and 70%, with one very well managed herd a little lower.' 21 Although Professor Morris is of the opinion that a new virus is the likely co-factor with PCV2, in causing PMWS, he accepts the possibility that the causal agent may be a new, particularly virulent, strain of PCV2. In his report, he used the term 'Agent X' to cover both these possibilities. He was not prepared to rule out the further possibility that Agent X operates in conjunction with other agents. He did, however, disagree with the opinion of some other researchers that the critical co-factor of PMWS was genetic or arose out of handling or environmental factors. He said those possibilities are excluded by the findings of a study in New Zealand. Apparently, PMWS has been encountered in a number of North Island piggeries, but nowhere on the South Island. Yet, according to Professor Morris, a detailed comparison between the two sets of piggeries revealed no significant difference between the genetic characteristics of the pigs, their handling or environmental factors. 22 I need not form any view about the likely cause of PMWS. The cause is a matter of vigorous debate between well-qualified scientists, without there being any conclusive proof of the correctness of any particular view. The legislation 23 The Quarantine Amendment Act 1999 (Cth) introduced into the Act what counsel for the respondent (Mr J Basten QC, Mr S J Gageler SC and Mr G R Kennett) described as 'three key concepts'. They described these concepts in this way: '(1) that quarantine is concerned not only with what occurs at the border but extends to encompass the "introduction, establishment or spread" of a disease: s 4(1)(b); (2) that quarantine measures may extend beyond simple prevention to "control": s 4(1)(b); (3) that there will be a "level of quarantine risk" which is a function not only of the probability of a disease being introduced, established or spread but also of the probable extent of any harm: [s]5D.' 24 Counsel referred to the following statement in the explanatory memorandum for the Bill which became the 1999 amending Act: 'Australia's quarantine policy is based on the concept of the management of risk to an acceptably low level. The natural and economic movement of people, animals, plants and goods results in an inevitable quarantine risk to Australia. Australia's approach is to manage risk in a manner that provides appropriate protection for Australia, is based on scientific reasoning and is consistent with international rules and standards.' 25 Section 4(1) of the Act is as follows: 'In this Act, quarantineincludes, but is not limited to, measures: (a) for, or in relation to: (i) the examination, exclusion, detention, observation, segregation, isolation, protection, treatment and regulation of vessels, installations, human beings, animals, plants or other goods or things; or (ii) the seizure and destruction of animals, plants, or other goods or things; or (iii) the destruction of premises comprising buildings or other structures when treatment of these premises is not practicable; and (b) having as their object the prevention or control of the introduction, establishment or spread of diseases or pests that will or could cause significant damage to human beings, animals, plants, other aspects of the environment or economic activities.' (Original highlighting) 26 Section 5 of the Act defines the term 'Quarantinable disease' as 'any disease declared by the Governor-General, by proclamation, to be a quarantinable disease'. PMWS has been proclaimed a quarantinable disease. 27 Section 13 of the Act empowers the Governor-General to do various things by proclamation. Pursuant to that power, the Governor-General made the Proclamation. It contains two relevant provisions. 28 Clause 39 of the Proclamation prohibits the importation into Australia of meat or meat product unless the Director has granted an importation permit. 29 Clause 70 of the Proclamation stipulates the matters that the Director must take into account in deciding whether to grant an importation permit. It relevantly says, the Director: '(a) must consider the level of quarantine risk if the permit were granted; and (b) must consider whether, if the permit were granted, the imposition of conditions on it would be necessary to limit the level of quarantine risk to one that is acceptably low; and … (c) may take into account anything else that he or she knows that is relevant.' Clause 70 contains a note that the term 'level of quarantine risk' is defined in s 5D of the Act. 30 Section 5D of the Act says: 'A reference in this Act to a level of quarantine riskis a reference to: (a) the probability of: (i) a disease or pest being introduced, established or spread in Australia, the Cocos Islands or Christmas Island; and (ii) the disease or pest causing harm to human beings, animals, plants, other aspects of the environment, or economic activities; and (b) the probable extent of the harm.' The import risk analysis report (i) Introductory 31 As the IRAR is at the heart of this case, it is desirable immediately to refer to the parts of it that are relevant to this case. 32 The IRAR itself comprises two volumes totalling 767 pages. A third volume contains annexes. The IRAR was the product of a lengthy inquiry undertaken by a five-member risk analysis panel ('the Panel') comprising: Dr David Banks, General Manager of Animal Biosecurity in Biosecurity Australia; Dr Robyn Martin, Manager of Animal Biosecurity in Biosecurity Australia; Dr Kevin Doyle, Veterinary Director of the Australian Veterinary Association; Dr Ross Cutler, Consultant Specialist Veterinarian; and Professor Colin Wilks, Consultant Microbiologist. Two technical working groups, one of which was concerned with PMWS, assisted the Panel. 33 The IRAR explained that, under existing policy, uncanned, uncooked pig meat could be imported only from the South Island of New Zealand, Canada and Denmark. The approximate proportion of imports from each of these sources was respectively 5%, 60% and 35%. Canadian and Danish pig meat had to be imported deboned and cooked on arrival in Australia. These requirements were imposed in order to address the quarantine risk associated with the potential presence of porcine reproductive and respiratory syndrome ('PRRS'). Pig meat could be imported from any country if the meat was imported in a sealed container, the contents of which had been heated to at least 100°C. 34 Although the evidence and arguments are confined to the risk to Australian pigs from PMWS, it is important to note that the IRAR also discusses the risk posed by 11 other diseases. 35 In an introductory section of the IRAR, the Panel referred to s 70 of the Proclamation and commented: 'As can be seen from the above extracts, the legislation establishes the concept of the level of biosecurity (quarantine) risk as the basis of decision-making under Australian quarantine legislation. Import risk analyses are a significant contribution to the information available to the Director of Animal and Plant Quarantine - a decision maker for the purposes of the Quarantine Proclamation. Import risk analysis is conducted within an administrative process - known as the IRA process (described in the IRA Handbook). The purpose of the IRA process is to deliver a policy recommendation to the Director of Animal and Plant Quarantine that is characterised by sound science and by transparency, fairness and consistency. The key elements of the IRA process are covered in "Import Risk Analysis" below.' 36 The IRAR proceeded to refer to Australia's international rights and obligations, stemming primarily from the World Trade Organization's Agreement on the Application of Sanitary and Phytosanitary Measures ('the SPS Agreement'). The Panel explained: 'The SPS Agreement recognises the right of WTO Member countries to determine the level of sanitary and phytosanitary protection they deem appropriate, and to take the necessary measures to achieve that protection. Sanitary (human and animal health) and phytosanitary (plant health) measures typically apply to trade in or movement of animal and plant based goods within or between countries. The SPS Agreement applies to measures that may directly or indirectly affect international trade and that protect human, animal or plant life or health from pests and diseases or a Member's territory from a pest.' 37 The Panel noted that the SPS Agreement allows WTO members to determine the appropriate level of sanitary and phytosanitary protection ('ALOP') but this determination is required to be based on scientific principles and not maintained without sufficient scientific evidence. The Panel went on to say that Australia expresses its ALOP in qualitative terms. The ALOP is 'aimed at reducing risk to a very low level, but not to zero'. 38 As I have mentioned, there is uncertainty as to the identity of the co-factor or co-factors which, with PCV2, cause the onset of PMWS. Some scientists question whether there is an infectious co-factor. However, the Panel proceeded on the conservative basis that this is at least a possibility, as all the relevant experts who gave evidence in this case agreed. Accordingly, the IRAR frequently referred to an 'infected' carcass or 'infected pigs', although the more neutral term 'affected' might better describe the present state of expert knowledge. (ii) The risk estimation matrix 39 The IRAR set out, as Table 10 on p 14, a 'Risk estimation matrix'. It took the following form: 40 At p 24, in a section of the IRAR explaining their adopted method of assessing import risk, the Panel said the assessment should take into account both 'the likelihood that a pathogenic agent will enter an importing country' ('release assessment') and 'the likelihood that susceptible animals will be exposed to that agent' ('exposure assessment'). In the context of PMWS, 'susceptible animals' means healthy pigs. PMWS has not been found in humans or in other animal species. The likelihood of establishment and spread, and the biological and economic consequences of introducing a pathogenic agent were to be determined through a 'consequence assessment'. The Panel said: 'The risk assessment for each identified agent concluded with "risk estimation", the combination of the likelihoods and consequences, and yielded the unrestricted risk estimate.' 41 The 'unrestricted risk' is the risk that exists before the application of conditions designed to reduce the risk level. The risk after application of recommended conditions is termed 'restricted risk'. 42 At p 27 of the IRAR, the Panel said: 'Quantitative data were not available to support many of the probabilities assigned to the pathway steps considered in this analysis'. 'Likelihoods' were therefore derived from 'expert judgements'. If the likelihood was described as 'high', this meant 'the event would be very likely to occur'. If it was 'moderate', this meant 'the event would occur with an even probability'. If it was 'low', 'the event would be unlikely to occur'. If it was 'very low', 'the event would be very unlikely to occur'; and so on to 'extremely low and negligible'. 43 The Panel said: 'In order to ensure consistency in the usage and interpretation of these six terms and definitions, and to provide a framework under which they could be logically and transparently combined, the 0-1 interval for likelihood was divided into six categories. Events considered almost certain to occur were assigned a likelihood of 1. High >0.7 ® 1 Moderate >0.3 ® 0.7 Low >0.05 ® 0.3 Very low >0.001 ® 0.05 Extremely low >10-6 ® 0.001 Negligible >0 ® 10-6' 44 This means that a likelihood is to be regarded as 'high' if there is a 70-100% probability of its occurrence, 'moderate' if the probability is 30-70% and 'low' if 5% to 30%. Counsel for the applicants criticised the wide range of probability included in 'moderate'. 45 The IRAR stated that '[t]he band of cells in Table 10 [reproduced at para 39] marked "very low risk" represents Australia's ALOP, or tolerance of loss'. These cells occur in the following situations: (i) 'very low' consequences, even with high or moderate likelihood of entry and exposure; (ii) 'low' consequences with low likelihood of entry and exposure; (iii) 'moderate' consequences with very low likelihood of entry and exposure; (iv) 'high' consequences with extremely low likelihood of entry and exposure; (v) 'extreme impact' consequences with negligible likelihood of entry and exposure. (iii) Assessment of consequences 46 The consequences line has an important bearing on the risk matrix results. So it is desirable to note the Panel's explanation of assessment of consequences. At p 63, the IRAR stated that direct and indirect consequences were estimated at four levels: (local, district/regional, State/Territory and national). The IRAR said the first step was to assess the magnitude of the impact of the particular disease on the national economy or Australian community; if there was no discernible impact at that level, in descending order the magnitude of impact at other levels would be investigated. 47 In assessing consequences, the Panel addressed exposure to PMWS of three different groups of Australian pigs: feral pigs, backyard pigs and pigs in small commercial piggeries. It seems the Panel had in mind the possibility that pigs in the latter two groups might be fed discarded pig meat scraps and that feral pigs might obtain them by scavenging at rubbish tips. 48 The Panel's discussion of consequences, in relation to PMWS, commences at p 385 of the IRAR. The technical information set out in the IRAR includes a comment that '[i]ncreasing evidence continues to support the hypothesis that PCV2 is essential for the development of PMWS'. Consequently, the Panel paid some attention to transmission of this disease. At p 389, the Panel said it was unaware of any studies that have examined skeletal muscle for the presence of PCV2 viral antigen or virus'. The Panel also said '[i]t is unknown if pigs can be infected orally with PCV2'. However, it said: 'The detection of the virus in oronasal secretions and faeces is compatible with an oral route of transmission'. 49 Pages 390-391 contain a 'release assessment' in which the Panel referred to a series of potentialities and ascribed a likelihood to each. They were as follows: R1 - 'the likelihood that a source herd is infected' - assessed as 'moderate'; R2 - 'the likelihood that a slaughter-age pig from an infected herd is infected' - assessed as 'moderate'; R3 - the likelihood of non-detection. The Panel thought the sensitivity of ante-mortem, slaughter and processing procedures in detecting and removing subclinically infected pigs was 'extremely low'. In other words, there was an extremely high likelihood that any subclinical PMWS infection in imported carcasses would escape detection; R4 - 'the likelihood that the pathogenic agent will be present in the meat harvested for export' - assessed as 'moderate'; R5 - 'the likelihood that the pathogenic agent will not be destroyed by the post-mortem decrease in muscle pH that accompanies carcass maturation - assessed as 'high'; R6 - 'the likelihood that the pathogenic agent will not be destroyed during cold storage and transport' - assessed as 'high'. 50 The Panel offered this conclusion: 'When these likelihoods were inserted into the simulation model, it was concluded that, in the absence of risk management and without considerations regarding the exporting country, there was a "low" likelihood that imported pig meat derived from an individual carcass would be infected.' (iv) Estimate of annual exposure 51 The Panel went on to consider annual exposure assessments ('the L factor') for each of the three groups of pigs. It is unnecessary to set out the components of each assessment. The Panel's conclusion in respect of each group of pigs was that the overall annual likelihood of entry and exposure for that group was 'high'. (v) Estimate of outbreak scenario likelihoods 52 The Panel then estimated the likelihood of each outbreak scenario. It assessed there was a moderate likelihood of transmission of the disease from an exposed herd of feral pigs to a more general population of feral pigs and to backyard pigs but only a low likelihood of transmission to a more general population of domestic pigs, including pigs in medium-large piggeries. However, with infected backyard pigs there was a high likelihood of transmission to pigs in a more general population of pigs. Similarly, if the infection was in small commercial piggeries. (vi) Estimate of impact of PMWS in various groups 53 In assessing the impact of an outbreak of PMWS in each of the three postulated groups, the Panel used letters (A to G) to indicate a rating. The earlier the alphabetical position of the letter, the less significance was to be ascribed to that effect. The significance to be ascribed to each of the letters depended on whether the assumed effect had consequences on a national, State/Territory, district/region or only local level. That appears from Table 8 of the IRAR which is reproduced below: 54 The Panel concluded (at p 399) that an outbreak of PMWS that was confined to a directly exposed group of pigs would have only a B rating on animal health and an A rating on eradication programs and trade and industry (because the presence of the disease would be unlikely to be detected). However, if the disease spread to a local population of pigs in backyard or small commercial piggeries, the effect on animal health and domestic trade would be C, although the effect on international trade only B. 55 The scenario assessment relating to a possible secondary spread, via feral pigs or other means, to a more general population of domestic pigs, including medium-large commercial piggeries, is both more important and more controversial. At p 402, the Panel explained: 'Under this scenario, PMWS would have established in a broader population of commercial piggeries (including medium-large piggeries) and be identified. If the disease was not widespread in the Australian pig population, a control program may be implemented, alternatively, if widespread, control would likely be left to individual producers.' 56 The Panel noted high mortality rates in the United Kingdom, although they were apparently lower in Canada, the United States and Germany. The Panel said: 'On balance, the direct impact on animal health was considered unlikely to be discernible at the national level, but would be of minor importance at the State level. This gave the disease a rating of "D" for this criterion.' 57 The effect on domestic trade was similarly assessed. No figures were mentioned. Nor did the Panel make any distinction between the possible effect of PMWS in one State or Territory, as compared with any other State or Territory; for example, because of differences in their pig populations. Table 8 allowed State/Territory effects (however serious) to be categorised only as 'minor' (impact score D) or 'unlikely to be discernible' (impact score C). Unlike the situation in relation to national effects, there was no option of classifying a State/Territory effect as 'highly significant' (impact score G) or 'significant' (impact score F). 58 In considering the indirect effect of a secondary spread of PMWS to a general population of domestic pigs, the Panel assigned a C to the impact on eradication, control and compensation strategies and a D to the effect on domestic trade or industry. 59 The Panel said that, when the direct and indirect impacts of PMWS in each scenario were combined, using the decision rules that it had adopted, the consequences ranged from 'negligible' to 'low'. Combining these consequences with the assessed likelihoods, the overall likely consequences associated with the exposure of feral pigs to infected pig meat scraps were considered very low, and those for backyard pigs and pigs in small commercial piggeries were considered to be low. 60 It is to be noted that this assessment was made on an annual basis; that is, there will be a 'low' risk to backyard pigs and pigs in small commercial piggeries in each year, not a low risk of infection overall. The Panel did not make any assessment of the degree of risk of infection over a longer period of time. (vii) Recommendations for reduction of risk 61 A 'low' risk exceeds the 'very low risk' that is Australia's ALOP. Consequently, in the concluding chapter of the IRAR, the Panel discussed the options for further reducing the risk. At page 745 it said: 'The likelihood that PMWS could enter, become established and/or spread in Australia via imported pig meat could, in theory be reduced by the application of some or all of the following measures: · a requirement that the pigs of origin had never been in a PMWS infected country or zone since birth; · a requirement that slaughter and processing ensured removal of organs and tissues which are sites of predilection for the virus; · reduction in the volume of pig meat waste discarded in Australia. As porcine circovirus has been reported to be stable at a temperature 70şC for 15 minutes, the Panel did not examine the direct effect of processing (other than canning) on the destruction of this virus. Options were examined to identify the least trade restrictive measures which would reduce risks within Australia's ALOP.' 62 In relation to canning, the Panel commented at page 745: 'Australia currently accepts shelf stable canned pig meat from any source country subject to certain conditions. The Australian import conditions for canned meat include a requirement that all portions of the contents have been heated to at least 100şC. Porcine circovirus has been reported to be stable at 70şC for 15 minutes. The Panel considered that PCV2 would be inactivated in canned pig meat heated to at least 100şC.' 63 The Panel then turned to the possibility of modified dressing of the carcass. I will set out its discussion of this matter, omitting publication references. Counsel for the applicants (Mr J T Gleeson SC and Mr M J Leeming) complained that the passage dealt only with PCV2 (this apparently being 'the virus' referred to) and assumed, without knowing, that what is true of PCV2 is also true of PMWS. The passage read: 'Modified dressing of the carcass to remove certain tissues where the virus has an affinity would influence the fourth step in the release pathway (R4). This step describes the likelihood that the pathogenic agent would be present in meat harvested for export. Modified dressing of the carcass may also influence the exposure step L2, which describes the likelihood that a waste unit would contain a sufficient dose of PCV2 to initiate infection. Lymphoid tissues are the primary target tissues of PCV2, as for many other viruses. Viral antigen or nucleic acid is found in lymphoid tissues, including peripheral lymph nodes of clinically healthy and diseased pigs. Viral nucleic acid has been detected in bone marrow … and in serum for up to 16 weeks ... Viral nucleic acid has also been detected in sera of slaughter-age pigs … It is possible that these pigs may have been recently infected. The Panel examined removal of major peripheral lymph nodes (i.e. removing the head and neck and removal of other major peripheral lymph nodes such as inguinal, popliteal, axillary etc) together with deboning the carcass on the above likelihoods. The Panel considered that removal of the head and neck, including any remaining tonsillar tissue, and lymph nodes draining the pharynx, other major peripheral lymph nodes and deboning could reduce the likelihood assigned to R4 from "moderate" to "low". Moreover, if meat is sourced from areas other than the head and neck and other major peripheral lymph nodes are removed together with bone, the amount of virus present in a waste unit would be reduced. It is known that the virus has a strong affinity for lymph nodes … Virus titres of approximately 105 to 106 TCID50/g of lymph node have been reported from clinically healthy pigs experimentally infected with PCV2 … Levels of virus in both serum and lymph nodes decrease with increasing time post-infection. In persistently infected pigs levels of virus in tissues are likely to be low. Given the likely level of virus in muscle per se, the composition of pig meat waste (bone-out) and the volume of waste consumed by a pig, it was considered that the likelihood assigned to L2 could be reduced from "moderate" to "very low".' 64 Turning to reduction in the volume of discarded waste, the Panel opined: 'that if meat was deboned and processed either by cooking or curing, the proportion of pig meat purchased by households and food service establishments that was discarded as waste would be reduced to one tenth of that estimated for the unrestricted risk.' 65 The Panel went on to express the opinion that a combination of removal of peripheral major lymph nodes, deboning and cooking or curing, would reduce the previously 'low' risk to 'very low' and thereby meet Australia's ALOP. 66 The Panel attached to the IRAR a draft biosecurity policy document setting out proposed quarantine requirements in respect of the importation of pig meat. They included (cl 3.2) a requirement that the Official Veterinarian certify, amongst other things, that: 'The pigs from which the meat was derived have been kept since birth in a country or zone which is recognised by Australian authorities as free from post-weaning multisystemic wasting syndrome (PMWS). or The pig meat has been processed by canning such that all portions of the contents have been heated to at least 100şC. or The meat has not been derived from the head or neck, major peripheral lymph nodes have been removed, meat has been deboned and the product is processed (cooked or cured). or The pigs from which the meat was derived are not from a country or zone recognised by Australia as free from PMWS and the meat has not been processed by cooking or curing as above. Note: In this case, the meat must be processed in Australia.' (Original emphasis; detailed instruction notes omitted.) 67 The applicants make no complaint about permissions for importations that satisfy the first or second of these four alternatives. They challenge the sufficiency of the third and fourth options. Other evidence (i) Introductory 68 It is not necessary to summarise all the other evidence put before the Court. A number of affidavits were read and much documentary evidence was tendered. Most of this evidence was uncontentious, except occasionally as to its relevance. The major factual issue raised by the affidavits concerns the cause of PMWS, a matter about which it is unnecessary for me to reach a conclusion. 69 However, there is also expert evidence relating to the methodology used in the IRAR, whose alleged deficiencies lie at the heart of the case. It is desirable for me to summarise the views of the experts who dealt with that subject. I will also refer to some documentary evidence tendered by the respondent concerning the circumstances surrounding the Permit Decision. (ii) The IRAR methodology (a) Professor A N Pettitt 70 Professor A N Pettitt is Head of the School of Mathematical Sciences at the Queensland University of Technology. He has published widely, especially in relation to theoretical and applied statistics. For some years, he was a co-editor of Biometrics, a refereed international journal publishing statistical theory related to biological sciences. Professor Pettitt provided a report to the applicants and made an affidavit that was filed on their behalf. 71 In section 2 of his report, Professor Pettitt noted that the methodology adopted by the IRAR consisted of six parts: (1) Evaluating and reporting likelihood; (2) Release assessment; (3) Exposure assessment; (4) Consequence assessment; (5) Risk estimation; and (6) Method of risk management. 72 Professor Pettit commented: 'Parts 1, 2 and 3 rely on a risk modelling approach which can be described as Monte Carlo simulation, as implemented in the software package @Risk, together with input values which are elicited from scientific and industry experts. In this report I examine methodological, modelling and implementation issues which are pertinent to all parts of the process. I find that the uncertainty in estimates of modelling inputs has been inadequately and misleadingly modelled and inappropriate assumptions have been made in the modelling. Additionally, the IRA Report comes to its conclusions based on annual risk whereas an assessment over a longer time period gives insight into the time until a significant disease event occurs which might better reflect Australia's [ALOP].' 73 Professor Pettitt went on to describe, in some detail, the evaluation process undertaken in each of the six parts. 74 In section 3 of his report, Professor Pettitt offered a critical overview of the IRAR's methodology. 75 Professor Pettitt noted that the IRAR 'describes a complex method of import (or quarantine) risk assessment where the method of risk estimation is inherently quantitative with risk varying directly as certain quantities vary'. He gave four examples of variable quantities: (a) the volume of uncooked pig meat that might be imported; (b) the amount of waste generated in a twelve month period; (c) the proportions of Australia's population residing in different regions; and (d) the proportions of imported pig meat likely to be purchased by establishments and households in Australia. 76 In recognition of these uncertain values, Professor Pettitt observed, the IRAR represented them by probability distributions, 'spreading uncertainty over a range of values with different weights'. He went on: 'Additionally risk varies with other quantities used in the IRA Report's methodology which are generally called likelihoods. A likelihood is an uncertain quantity but specifically refers to the probability of a specific event occurring and is mathematically restricted to values in the range 0 to 1. Examples of likelihoods in the IRA Report include the following: · the likelihood that a source herd is infected (R1) and the five other likelihoods (R1 to R6) for the steps described in the release scenario for imported pig meat …; · the three disease specific likelihoods (L1, L2, L3) referring to waste units, and the six likelihoods (L4S and L5S) concerning accessibility of and location by feral pigs to waste units …; and · three likelihoods in each of Table 6 and Table 7 … Such likelihoods are assessed in the IRA Report … using a qualitative approach using a six value ordered scale (from "high" to "negligible") which is imprecise and ambiguous and therefore would appear to ignore some of the quantitative evidence in the IRA Report and the assumed expert judgement of the IRA Team. However, such determined likelihoods are combined through a series of mathematical formulae and steps, … to determine an annual likelihood of entry and exposure. The combination is effected using an assumption which replaces a qualitative likelihood value by a stochastic variable which takes with equal probability any value in a range prescribed by the qualitative value of the likelihood … For example each likelihood assigned a value of "moderate" is assigned with equal probability any value in the range from 0.3 to 0.7. In practice in the IRA Report this is effected by Monte Carlo simulation using a software package called @Risk. Generally, a particular likelihood could be quantitatively assessed by experts to give a best estimate and a range of probable values leading to a probability distribution representing this expert judgement, in the same way that the annual volume of uncooked pig meat that might be imported is quantitatively assessed in the IRA Report. The methodology of import risk assessment adopted in the IRA Report therefore has the following consequences for risk estimation: · expert judgement of likelihoods is not represented accurately and spurious uncertainty is introduced through @Risk simulations; and · more generally uncertainty of expert judgement and assessment is not handled in a consistent and accurate manner in the IRA Report. Thus, the conclusions of the IRA Report as to risk are unsound.' 77 Professor Pettitt pointed out that consequences are evaluated in the IRAR in a qualitative manner. Any uncertainty in the qualitative assessment is ignored. Professor Pettitt adopted a comment in a publication of the Food and Agriculture Organization of the United Nations about uncertainties in estimating the probability of introduction of a pest and its economic consequences. The comment said: 'It is important to document the areas of uncertainty in the assessment'. Professor Pettitt went on: 'Certainly consequences, being the impacts of a disease, can be analysed for their possible economic effects and the IRA Report's methodology states at page 63 this is a consideration. Such an assessment should be quantitative and in monetary terms … In addition, the IRA Report states that "the consequences are mutually exclusive" … so that total losses are disaggregated over various direct and indirect effects and cannot be counted twice. Qualitative impact scores are introduced to assess the consequences of each outbreak scenario for each exposure group, but these are qualitative in nature. These impact scores are then combined over direct and indirect impacts using a set of rules which is not additive, whereas they should be additive if economic losses were being combined to find the total loss … Likely consequences for each outbreak scenario are then obtained as a combination of the likelihood that the scenario would occur (being the likelihood of establishment and/or spread for the outbreak scenario) and the consequences associated with that outbreak scenario using a matrix of rules described in Table 9 on page 68 of the IRA Report. Both scales and the result of the combination are qualitative which means that the rules appear arbitrary depending on ambiguously defined scales of measurement. The likely consequences for each of the outbreak scenarios for each exposure group are then combined using a set of rules which are not additive in nature. It would be appropriate to have an additive set of rules for likely consequences measured in monetary terms, but the IRA Report's rules appear to follow an arbitrary "order of magnitude" set of rules … Overall likely consequence assessment appears to be based on arbitrary rules which do not follow an additive basis which would be appropriate for assessing economic, social and environmental losses in monetary terms. Such an approach leads to inaccuracies in the assessment of likely consequences and therefore risk assessment in the IRA Report.' (Original emphasis) 78 In dealing with risk estimation, Professor Pettitt observed as follows: 'Partial annual risk of exposure for each exposure group is assessed by a combination of likelihood of entry and exposure and consequences of entry, establishment or spread … Likelihood of entry and exposure is assessed on a qualitative scale as a result of @Risk simulation values being converted from a probability on a quantitative scale, over the range from 0 to 1, to a single value on a qualitative scale of six ordered values (being "high" to "negligible"), ignoring any uncertainty in likelihood estimation if it had been accurately assessed and also possibly losing accuracy of the likelihood estimate. Consequences of entry, establishment or spread are also assessed on a qualitative scale of six ordered values (being "extreme" to "negligible" impacts) ignoring any quantification of losses. Mathematically, risk is the multiplication of probability (likelihood) and loss (consequence) but Table 10 … of the IRA Report used for this combination does not display characteristics of this mathematical operation. For this to be possible, numerical values have to be assigned to the consequence values but that has not been done in the IRA Report. Partial annual risks for each of the exposure groups are combined to give the overall annual risk for a disease using a set of rules which again are not additive but largely arbitrary … The rules appear to follow an "order of magnitude" rule. The IRA Report's approach to risk estimation therefore loses accuracy by using qualitative scales to measure likelihoods and consequences and arbitrary rules to combine these quantities. It also fails to consider any uncertainty in the assessment of likelihoods and consequences and therefore gives a false impression of precision in the estimation of overall annual risk.' 79 Professor Pettitt also criticised the fact that likelihood and consequences are assessed on an annual basis. He thought it is 'reasonable to consider longer periods of time to properly assess import risk'. 80 Professor Pettitt concluded section 3 of his report by saying: 'It would appear that the IRA Report, through consideration of unrestricted and restricted risks that meet Australia's ALOP, gives conditions under which imports of pig meat can take place with minimal trade restrictions. It does not give a basis for decision making which takes into account: (a) the appropriate assessment of uncertainty for likelihood estimation and the uncertainty of consequences assessment; (b) the appropriate assessment of consequences and overall annual risk in monetary terms; or (c) consequences of any realistic scenarios not considered in the IRA Report.' 81 The points made by Professor Pettitt were developed in some detail in the remainder of his report. It is not necessary to set out that detail. Professor Pettitt commented, at para 4.1, that 'the literature on risk analysis describes two methods for risk assessment … the qualitative method and the quantitative method', although he noted that a more recent method (semi-quantitative risk assessment) has been described by Mr David Vose, who gave evidence in this case for the respondents. 82 Apparently Professor Pettitt had no objection to an approach that adopted either the qualitative or quantitative method. He did have a problem about a mix of the two. At para 4.1.6 he said: 'The IRA Report … uses what it describes as a semi-quantitative method to assess likelihoods and proportions and to assess risks and restricted risks. However the approach of the IRA Report would appear to use all three methods, qualitative, semi-quantitative and quantitative, in an ad hoc mixed method which leads to inaccuracies. For example, when the method used in the IRA Report is considered for likelihood assessment, then expert judgement and its uncertainty for unknown quantities is constrained in unnecessary ways and therefore generally misrepresents that judgement. For example, consider the assessment of R1 the likelihood that a source herd is infected for Post-weaning multisystemic wasting syndrome … Evidence presented gives values of 18% to 20%, 5.6%, over 20%, over half. The assessment is given as "moderate" covering the range 30% to 70% whereas the data suggest a range of 5% to over 50%.' 83 During the course of his oral evidence, Professor Pettitt clarified the meaning he attributed to 'qualitative', 'quantitative' and 'semi-quantitative' in this context. He said: 'If it is purely quantitative you can assign numbers to every ingredient in the analysis and you base that on the best number you've got and then you just do your arithmetic? --- Yes. Right. Now if it is qualitative it no doubt takes on board whatever factual material you've got and then makes assessments as to the significance of that actual material. Those assessments depending on the individual participants' expertise and degree of intelligence, I suppose, common sense and so on, is that right? --- That's right, yes. And when you've got semi quantitative, you've got some data put in but judgments are being made along the way, is that why you use that phrase? --- Yes because in all of the methods I suppose there is a mix of data, hard data being used, and what we might call expert judgment that has to interpret that data in the context of in here the IRA. In some instances there will be more data. In other instances there will be less data and therefore depend more on expert judgment.' 84 Professor Pettitt complained of unnecessary uncertainty in the IRAR; for example, the number of Australian households was stated to be within a range of 6.8 - 7.6 million, without regard to available Australian Bureau of Statistics ('ABS') information on this subject. He said: 'The approach used in the IRA Report therefore generally does not represent the expert judgement and its uncertainty in an accurate manner and leads to unreliable calculations of the overall annual risk for the diseases under consideration and, therefore, unreliable conclusions about whether or not Australia's ALOP is met for a particular disease. These inadequacies are carried through the Monte Carlo simulations which are conducted using the @Risk software package, and result in an arbitrary output distribution which does not adequately represent the underlying science, nor the uncertainty about model inputs in judgments elicited from experts. Monte Carlo simulation allows complicated mathematical calculations to be carried out by drawing random numbers to approximate the calculations to any desired degree of accuracy. The IRA Report's estimation of the annual likelihood of entry and exposure is a complicated mathematical calculation and the @Risk output distribution incorporates the uncertainty implied by the IRA's methodology for the input likelihoods. The necessary consequence of each of these deficiencies is that the IRA Report does not provide a sound and reliable basis for overall risk assessment for each of the diseases assessed using the methodology described in it. 85 At para 6.2 of his report, Professor Pettitt referred to the Panel's choice of the 50th percentile for determining likelihoods. He said: 'It is standard statistical practice when eliciting likelihoods from experts to also elicit an indication of the uncertainty to be attached to each numerical value, typically as a probability distribution. [He cited some references for this statement.] This uncertainty should be meaningful, expressing as appropriate natural variability and/or lack of precise knowledge. The uncertainty should be based on the state of scientific and expert knowledge and opinion, rather than arbitrarily assigned according to the category of the estimated likelihood as pages 27 to 28 of the IRA Report indicate that they should be. The IRA Report (at page 27) appears to argue that under "Modelling qualitative expert judgement" that if likelihoods use expert judgement then quantitative probability distributions cannot be used which appears to contradict the cited references. The consequence for the modelling is that there will be an incorrect variability of results over the @Risk simulations, which will not correctly represent the state of scientific and expert knowledge. Then characterisations made on percentiles, such as the 50th or 95th, of this incorrect distribution will be invalid. Additionally, they would be largely arbitrary if they were to be compared with the distribution resulting from the uncertainty being correctly assessed for the expert judgment. In conclusion, the 50th percentile of @Risk simulations carried out using the methodology of the IRA Report does not provide a robust estimate of likelihood and is instead invalid and largely arbitrary.' 86 In a report filed on behalf of the respondent, Mr Vose defended the Panel's choice of the 50th percentile. Professor Pettitt responded to that defence in para 2.1 of a supplementary report. He said: 'Choice of the 95th percentile rather than the 50th expresses an appropriate degree of caution or conservatism. If we consider the likelihood of an event and we consider making a decision if this likelihood is less than a given threshold, then use of the 50th percentile (that is, is the 50th percentile equal to the threshold) implies that we have 50:50 degree of belief that the likelihood is less than the threshold, which we have set for making that decision, rather than greater, whereas use of the 95th implies we have a 95:5 or, equivalently, 19:1 degree of belief that the likelihood is less than the threshold rather than not. That is, in the latter case with the 95th percentile, we are much surer, in terms of our degree of belief, of being below the threshold and this reflects a more conservative approach to decision making. In the case of the IRA, an example would be the decision to classify a likelihood value, say L1, the likelihood that a waste unit is infected (see IRA pages 48 and 51), as either "low" or "moderate", say, where the threshold value is 0.3. Use of the 95th percentile reflects an appropriate conservative approach.' 87 Mr Basten took up this issue in cross-examination of Professor Pettitt. Professor Pettitt agreed that the choice of percentile reflects the person's level of confidence in the result. He said choice of the 95th percentile 'establishes a very high level of confidence'. He agreed it 'offers a cautious approach in terms of the interpretation of the output of the risk simulations'. 88 Professor Pettitt agreed that he favoured use of the 95th percentile in a situation such as this, and that he had described the Panel's choice of the 50th percentile as 'arbitrary'. He explained: 'To some extent the choice of any particular percentile is going to be one of judgment and then to some extent if one hasn't explained that judgment appropriately then it would be arbitrary.' 89 Mr Basten asked Professor Pettitt about his criticism of the Panel having assessed risk only over a one-year period. Professor Pettitt agreed it was impossible to speculate about future knowledge; one had to work on the basis of 'today's knowledge'. He explained: 'I think what is being attempted here with a longer period of time is just to imagine that if you were looking at a 10 year period of time then those 10 years would be like the first year. Nothing had changed much. So you would have 10 years similar to the first year and trying to assess what would happen over that period of time. Not taking into account any other what might be called developments going on in disease protection. Just trying to see the effect of the one year calculation being put over the longer period of time. There is nothing wrong with the one year period of time. … I am just saying that in terms of making decisions then it is useful to know what would happen, what was likely to happen over, say, a five or 10 year period of time. Extrapolating out the results of a one year period of time, keeping everything else essentially fixed. In broad terms we know, of course, what the answer is: the longer the period the closer you will get to certainty? --- Yes. Yes that is the nature of the IRA calculation.' 90 Professor Pettitt was asked about his criticisms of uncertainties in the IRAR. Mr Basten took him through three examples in which Professor Pettitt complained of the unexplained adoption of a figure, or range of figures, when ABS figures would have been available. The evidence went on: 'These were examples, I think you repeat at least one of them later, but these are examples of cases where there is data available? --- Well, I would have thought that the ABS would have been able to offer an opinion as to the accuracy of those numbers. I understand that but these are cases where there is numerical data indeed available as compared with the situation where there really is no specific data and one is making as it were a purely evaluative judgment on whatever science-based material you have. What I was going to ask you was that that is the situation where you criticise the use of a simple uniform distribution I think, is that right? --- In the semi-quantitative assessment of likelihoods, yes. I was making in the report in general comments as to how uncertainty is addressed in the IRA report. The idea of a simple uniform distribution is that you can't assign particular value to a number A or B which is higher or C which is lower, is that fair? --- Well, that's within the range, yes. So if you take uniform over 30 per cent 70 per cent then you're giving the same weight to all values between 30 and 70 per cent and no weight to values between nought and 30 per cent and 70 per cent --- Above 70? --- and 100 per cent. That may or may not be the best way to assess matters even where there is no specific accurate information, is that your point? --- That is my point. There is a process called elicitation which is well documented in the literature which is this trying to quantify in probability terms what the person, an expert, knows about an unknown quantity.' 91 Professor Pettit commented on the fact that the same weight was given to a probability lying anywhere between 30 and 70%, but significantly less weight to 29.9%. His point, as I followed it, was that the adoption of categories of likelihood, rather than the use of percentages, necessarily meant the adoption of arbitrarily chosen demarcation points. The professor made a similar point later, when Mr Basten was asking him about his comment that the proliferation of steps necessarily meant a reduction in the assessed risk; even if the risk was assessed as 'high' the step would be factored into the computer calculation by a series of figures whose mid-point was 85%. The more times one multiplies a given figure by 85%, the lower the result. Professor Pettitt said: 'Well I think the point I am making here is essentially the one that because of the semi-quantitative nature of the likelihood assessment then with each new step in a pathway and unless certain is used for that - certain likelihood value is used for that step then the implication is that the highest likelihood assessment is high, which is 85 per cent is its mid-point. So each new stat will bring the overall likelihood down by that proportion, 85 per cent. That is because the IRA and the experts are forced to use the semi-quantitative. They appear to be forced, they don't do anything else but use the semi quantitative approach. They do use it; yes? --- Therefore, rate things as high, or moderate. Rather than saying, my best estimate for that particular step might be 99 per cent or 98 per cent. Rather than the implied estimate which comes from high, which is 85 per cent. So they are not allowed to say, 98 per cent. It has to be their best estimate. A 98th per cent for a step then it would be categorised as high and converted to an average or mid-point value at 85 per cent. But that is assuming, isn't it, that they are relatively certain about one step? --- All I am saying is, their best estimate. I would have thought your experts - the experts in the IRA process would have a best estimate of each of these likelihoods and then there would be some uncertainty about that estimate. HIS HONOUR: Can I just check the arithmetic? At 8.1.13 you talk about an average equal to 0.85 or 0.52. You get that by multiplying 0.85 by 0.85 by 0.85 by 0.85 - four times? --- I hope so. But even if you were certain of something, so that you had recorded, well if not 100 per cent 99.9, you have still got to bring it down to 85? --- Yes. That is your point? --- That is the point of the criticism. Well, no, the IRA process allows certain so probably a few at 99.5 per cent certainly would push it up to one. But if it was 95 or 97 then it might be converted down to high and the 85 per cent. It is not clear what would happen. You could have four things, four steps, each of which you had a very high degree of confidence in, although not total certainty. The multiplication process obliges you to say it is only just marginally over half and therefore moderate? --- That is the consequence of using the semi-quantitative method. MR BASTEN: But your criticism does depend on a belief that in one of these hypothetic examples rather than saying, 85 per cent the scientist is willing to say, with a sufficient degree of certainty, 98 or whatever it may be? --- Well I am not clear how close to one or 100 per cent in an IRA process one would push it up to one and be certain. The implication might be that if one was 98 or 97 per cent then it would be categorised as high. As high? --- As high. Yes, it could be. I don't - it is not clear what goes to one and what goes to high in those circumstances. But that is part of the evaluative judgment; isn't it? --- Yes, it would be.' (b) Professor Ian Gardner 92 Professor Ian Gardner is a graduate in veterinary science from the University of Sydney. He is currently Professor of Epidemiology in the School of Veterinary Medicine at the University of California, Davis. He is a member (by examination) of the Australian College of Veterinary Scientists in Pig Medicine. He teaches Swine Herd Health to third-year veterinary students and is clinical veterinarian for the UC Davis Annual Science Swine Unit. Professor Gardner has published, and undertaken consultancies, concerning pig health and risk analysis. Professor Gardner gave evidence on behalf of the respondent. 93 In his report, Professor Gardner noted a request that he provide his opinion 'as to whether the release assessment (pages 390-391), the exposure assessment (pages 392-395) and the consequences assessment (pages 395-405) for PMWS in the Final IRA provide an appropriate basis for assessing: (a) the probability of PMWS being introduced into Australia and causing harm to pigs in Australia, in light of the risk management measures proposed at pages 745-749 of the Final IRA; and (b) the probable extent of that harm.' 94 Professor Gardner responded: 'In my opinion, the release assessment, exposure assessment and consequence assessment for PMWS in the final IRA adequately assess points 5(a) and 5(b), listed above. In many of the scenarios and component steps in the IRA, the Panel has taken a very conservative approach to the estimation of risk and its consequences. I believe that the final IRA is a fair and reasonable attempt to incorporate available information on PMWS and most importantly, I could find no "fatal" or "serious" flaws in the IRA. I define "fatal" and "serious" flaws as flaws that would have resulted in a substantial underestimation of probabilities and consequences of PMWS, or overestimation of the effects of the risk management measures. The release assessment provides a description of 6 sequential steps that need to be met for imported pig meat, derived from an individual carcass, to be infective. I believe the estimation is conservative because pigs may clear PCV2 infection as they age, i.e. the proportion of infected pigs consigned for slaughter is likely less than the proportion of pigs that ever became infected in the same herd. Estimation of the likelihood that source herds are infected (R1) was a difficult task for the Panel given the non equivalence of PMWS and PCV2, and the lack of high-quality international data. Inadequate data quality is most likely attributable to among-country differences in PMWS surveillance activity and lack of a standardized case definition for international recording of PMWS occurrence (which is not mandatory). Use of PMWS data alone is likely to underestimate the prevalence in pigs at slaughter, so the Panel's approach of using data for PCV2 nucleic acid in sera (page 390; release assessment) is a very conservative approach since not all the nucleic acid might be infective. In my opinion, accounting for the variability in prevalence of PCV2 among herds would have been a nice refinement to the model. On the other hand, the Panel seems to have taken a conservative stance by considering PCV2 prevalence to be moderate in slaughter-aged pigs. With regard to R3, it is likely that the sensitivity of lesion detection would vary with the severity of disease and clinical signs. The assumption of extremely low sensitivity is reasonable given that most pigs would be sub-clinically infected. … The IRA allows for different scenarios involving feral pigs (remote regions, rural regions, and large towns). In my opinion, the "moderate" and "high" likelihood assessed for feral pigs in remote and rural regions, backyard pigs and small commercial piggeries is a conservative estimate. The consequence assessment was based on the description of the characterisation of discrete outbreak scenarios for the 3 different exposure groups, i.e. feral pigs, backyard pigs and small commercial piggeries. Direct impacts on animal life or health and the environment were considered in addition to a range of indirect impacts on a variety of outcomes including trade. Given my understanding of pig production in Australia, I believe that the assessment covered the most likely exposure groups and direct and impact [sic: indirect] impacts should PMWS be established. Assessment of the overall impact of PMWS included scenario 4 - secondary spread to a more general population of domestic pigs (including medium-large commercial piggeries). This essentially considers an "endemic state" where PMWS is widespread in the industry. Given the awareness of the commercial pig industry and diagnostic laboratories, the existence of rigorous biosecurity practices, I believe that this scenario is unlikely. The likely consequences of the remaining 3 scenarios were all "negligible" or "very low". Inclusion of scenario 4 added a "low consequence" which meant that the overall consequences were "low" when all 4 possibilities were considered together. The combination of likelihood and consequences into likely consequences resulted in a "very low" overall estimate. Because I consider scenario 4 to be unlikely, I believe that this overall estimate is a conservative one. The 2003 Draft Guidelines (pp 48-50) describe use of expert judgements/opinion in risk assessment but does not include how to include expert opinion when experts disagree. A number of strategies exist and choice of different methods might result in different inferences. Ideally, how disparate views of the Panel members or the PMWS Technical Working Group were resolved should be documented. I assume the values considered in the model were a consensus view of the Panel (after having access to the same data) and that Panel members were in close agreement on all variables. It would have been helpful to include a formal sensitivity analysis for each probability or consequence to determine how robust the final conclusions of the IRA were to changes in such values. However, the Panel seems to have taken a consistently conservative approach to estimation, erring on the high side for both probabilities and consequences.' 95 In a later part of his report, Professor Gardner commented on the Panel's treatment of uncertainties. He said: 'The IRA has captured uncertainty in many of the input parameters by allowing for uniform distributions within probability categories. In a stochastic simulation, all values within the range would have an equally likely probability of being sampled. More informative distributions within these ranges could have been used, however there are clearly no data to make them more informative, e.g. use of beta or betapert distributions within each category. Accordingly, I disagree with Dr Morris' comment 6.3.6., in which he suggests that "no proper consideration is given to the issue of . uncertainty in the likelihood of PMWS ." In my opinion, the number of sequential steps in the likelihood should be determined by the biology/ecology of the pathogen under consideration and whether data are available for each of those steps … If one step is added at the end of chain with 5 steps, then the final likelihood for a 6-step process will always be less than the 5-step likelihood, unless that probability for the final step is 1. However, if a step is added in the middle of the process for which the starting and ending probabilities are known, then the final likelihood should remain unchanged. It is possible to include probability values of 1 for these intermediate steps, in which case the overall probability value would not change. It is common in IRAs to use "average values" for such analyses because tracking individual carcasses through the slaughter process and relating test results on a carcass by carcass basis is difficult and rarely done. I am not aware of any longitudinal data with respect to PCV2 that could have been used in the IRA. If these are available, the reviewer should provide references for the Panel.' 96 Professor Gardner concluded: 'My overall conclusion is that the findings of the final IRA are appropriate based on my review of the supportive documentation. In my opinion, the final IRA is a fair and reasonable attempt to incorporate published information about PMWS and its associated risks. The final IRA follows designated Biosecurity Australia guidelines as laid out in 2003 Draft Guidelines and also it appears to meet guidelines in the OIE code and the SPS Agreement under WTO. In general, I consider that the IRA was based on sound scientific principles using available knowledge, and that it is structured and transparent. In addition, I consider that it provides an outcome that 1) supports the estimation of "risk" and 2) enables risk to be evaluated against Australia's ALOP, and 3) allows the valid comparison of risk management measures. For each scenario, the IRA includes disaggregation of the likelihood estimate into its individual component likelihoods. I did not identify flaws that were sufficiently serious or fatal to negate the conclusions and recommendations in the IRA. Annex C documents the extensive consultation with national and international stakeholders and demonstrates that the Panel made a concerted attempt to address concerns that were raised by these outside groups. As is true for all IRAs, it is important to acknowledge any inherent flaws. These flaws can be broadly categorised as those pertaining to existing knowledge and data (outside of the control of the Panel) and those pertaining to the methodology in the IRA. For PMWS, the lack of a perfect "case definition (and differing opinions about its causes) and incomplete data made it difficult for the risk assessors to select the most appropriate data for risk estimation. The Panel clearly considered the involvement of unknown agents in the PMWS syndrome. The formal incorporation of a sensitivity analysis and documentation of how differences in expert opinion were resolved would have improved the final IRA but I don't consider that their omission has compromised the overall validity of the IRA. To set the final IRA in context, one should consider the fact that there have been 14 years of introduction of pig meat from Canada and Denmark (both countries have PMWS) into Australia with less restrictive import quarantine conditions and there have been no PMWS diagnoses in Australia. In future, it is possible that the volume of imported pig meat might increase compared with the last 14 years. However, the application of the risk management measures (specifically, reducing the amount of waste discarded, removal of the head and neck, bone and major lymph nodes prior to the export of pig meat for cooking/curing in Australia) which was not historically practised should only further protect the health of the Australian pig industry.' 97 During cross-examination by Mr Gleeson, Professor Gardner expressed agreement with a statement of Mr Vose that 'semiquantitative risk assessment techniques are currently not widely accepted in international risk issues because of difficulty in retaining transparency and because the process is open to abuse'. However, he also agreed with a rider by Mr Vose that 'semiquantitative risk assessment when properly executed is a transparent approach that supports the efficient management of a portfolio of risk issues without requiring complete quantifications to the risk or excess risk avoidance'. In relation to transparency, Professor Gardner said he had not seen the spreadsheets underlying the IRAR or any explanation of the Panel's reasoning for their selection of 'moderate' in relation to the likelihood that a source herd is infected. Neither had he seen any report of the number of times the Monte Carlo random allocation had been applied to the various R factors. He accepted failure to report that matter constituted a lack of transparency. 98 Professor Gardner said he had not seen a 95th percentile report about the risks discussed in the IRAR. He agreed it was part of his regular practice, when doing a risk analysis, to report on both the 50th percentile and the 95th percentile. 99 Professor Gardner was taken to a footnote (footnote 50 at p 98) in a document published in September 2001 by the Commonwealth Department of Agriculture, Fisheries and Forestry, Guidelines for Import Risk Analysis. The statement was: 'As a rule, it is recommended that the 95th percentile of an output distribution be reported. This conservative policy is based on a recognition that all models are (at last to some extent) imperfect representations of reality.' 100 Professor Gardner said: 'Yes, I would agree with footnote 50 on the proviso that the original distribution of the data is not shifted upwards. As has been done in this particular risk assessment where a very high artificial initial value was put on for the probability. But if I had modelled this differently with a midline value of 20 per cent I would probably have put forward the 95 per cent probability value as well. So, does this summarise your opinion? If, contrary to what you have just said, each of the individual integers was done in a fair and reasonable manner then you would have expected both the 50 per cent and the 95th percentile to be reported? --- If I was doing the risk assessment I would have modelled the things integer values and done 50th, 75th and probably 95th percentile. And that is standard practice; isn't it? --- That is common practice. Now when you read this report did a question occur to you along these lines: why can't I find in this report information at the 95th percentile; did that occur to you? --- No it didn't occur to me at the time.' 101 Professor Gardner accepted that the IRAR contained no explanation of its omission of 95th percentile reports. He said he had an impression that there was reasoning supporting the non-reporting of the 95th percentile in a document issued in 2003. He was vague about that; he said he would have to check. No such document was tendered in evidence. 102 Professor Gardner agreed with Mr Gleeson that steps R3, R5 and R6 in the release assessment can, for practical purposes, be put aside; the critical steps were R1, R2 and R4, all of which were assessed as 'moderate'. The professor did not agree that these elements could just as readily have been considered as one; he thought it was an 'aid to thinking' to consider each step separately. 103 Mr Gleeson then referred Professor Gardner to the IRAR's conclusion about release assessment: there was an overall low likelihood that imported pig meat derived from an individual carcass would be infected. The professor agreed the IRAR did not disclose what effect the recommended conditions would have on this likelihood. This flowed through to a similar omission in respect of L1 - the likelihood that a waste unit is infected. The professor's evidence went on: 'What I'm putting to you is that you agree that under this report which is the only information you had, you're not told after the measures what is the total R figure or the L1, do you agree? --- Yes, I think that's - that would be a correct interpretation and I find it - - - And I want to suggest to you that that is a fundamental lack of transparency in this report? --- It would have been helpful to have that documentation, that's correct. Do you agree with me? --- Yes. Thank you. Now, can you look at L1 on page 392 and this seems to be an assumption by definition that the likelihood that any individual waste unit which might get before a pig in a tip for instance being infected is the same likelihood as the source carcass had of being infected, correct? --- That's correct. Now, I want to suggest that in this report when you read it, you did not find any reasoning exposed as to why such an assumption should be made? --- The reasoning could have been clearer, I agree. What I want to suggest is, leaving aside views you might have on the topic and views others may have, when you read the report, you couldn't find exposed reasoning to justify the conclusion here expressed about L1, do you agree? --- No, it's not explicit. No, and that's a question on which different views might well be taken, correct? --- Correct. One view which might be taken is that if for example the pig meat is going into Australia cooked and cured it will largely end up in small goods factories, correct? --- Correct. In small goods factories there will be a significant risk that meat from different carcasses will be combined? --- I couldn't comment explicitly on how they are combined. That is outside my area of expertise. But certainly if you were trying to rationally assess L1 you would have to take into account through some intelligent process whether in the small goods factories multiple carcasses get combined, because if they did that might give you an increased L1, correct? --- It may give you an increased L1 under the proviso that infected and noninfected material is being aggregated in similar amounts, that is a possibility. But I think you agreed that in this report when you read it, you couldn't find any reasoning one way or the other on that type of issue? --- No, there is no obvious reason presented. But that's a topic about which various people hold opinions and there can be intelligent discussion about it, yes? --- Certainly. Now would you agree with me that the failure to explain the equation in L1 is a lack of transparency? --- It was presented as an assumption in the report. The basis of that assumption certainly wasn't totally clear.' 104 Mr Gleeson took Professor Gardner to the IRAR's conclusion that the L factor for feral pigs was 'high'; that is, within the range 70 to 100 per cent. The evidence continued: 'Wouldn't you want to know where it was in that range? --- In terms of the multiplication process if you take a mid line value that will practically end up as the 50th percentile which is how it is being operated from. So in fact from practical purposes it has been treated as being in the middle of the range, 85 percent. What I want to suggest is, you would want to know if the information is available whether it was in fact at the bottom of the range, the middle, or the top? --- If there were hard data available and it could be modelled as anything but a uniform distribution between 70 and 100 percent then I would certainly model it with a different shaped distribution if that were more appropriate.' 105 The professor's evidence went on: 'If you were told for example that the percentage for small commercial pigs was in fact a mean of 99.2 percent, that would effect your subsequent modelling wouldn't it? --- If I was told that and the data was convincing, yes. And you would no longer proceed simply to say it is somewhere between 70 and 100? --- Right, but on the other hand if it was 71 percent I would actually take that into account as well, at the lower end.' 106 Mr Gleeson showed Professor Gardner the Panel's working papers summarising the results of its computer modelling. These papers show a figure of 0.9993808 as the mean value of the exposure to infection of feral pigs and 0.9920131 for pigs in small commercial piggeries. These mean values each translate to a chance exceeding 99%. Mr Gleeson noted the IRAR simply assessed the exposure for each of these groups of pigs as 'high'. 107 Professor Gardner said he would like to have seen a median value but he accepted it would be about 99.0%, 'a high degree of certainty'. The evidence went on: 'But I think without taking much time about it you would agree that had you known that instead of modelling this as something between 70 to 100 percent of uniform distribution, you would have modelled this at the probability of one? --- I wouldn't model it as a probability of one. There is nothing certain in life. I would allow for some uncertainty below one. But I would certainly give it a high value. And you wouldn't have proceeded the way this report has proceeded, would you? --- I'd have to see the basis for all these numeric calculations, but if I was just presented with that single picture, if that was based on sound scientific evidence, I would say that is close to certainty. I would just like to suggest to you that this is an illustration that absolutely fundamental information necessary to be able to assess the rationality of this semi quantitative method was not contained in the report, do you agree? --- No, I don't totally agree with that. Well how much of that do you agree with? --- Well within the constraints of the categories presented, 70 to 100 percent, it certainly doesn't differentiate between any value within that range. That is absolutely a given. If you divided that value up and add another category, say 95 to 100 percent or 99 to 100 percent as representing something certain that would actually allow for capturing that type of output, but that wasn't provided. Certainly in this particular case it was 70 to 100 percent. If you had been doing this you wouldn't simply have reported this as a uniform distribution between 70 and 100 would you, knowing what's in front of you there? --- If I saw those data I would certainly have modelled it differently.' 108 Mr Gleeson questioned Professor Gardner about his view that the IRAR approach was conservative: 'I would like to put to you that before you gave your evidence today you were of the view that although the 95 percent probably should have been recorded it may be that conservative aspects of the model balance it all out, is that a fair assessment? --- I think that's a fair assessment, yes. I would like to put to you that having been through some of these matters today, including the very one we are looking at, your opinion that this is a model bearing on conservatism and therefore entitled to ignore the 95th percentile is one you no longer hold? --- No, that's not true. If I was doing the model, I would go back and maybe construct it differently, where I didn't shift the values to a very conservative range, in which situation I may well have considered the 95th percentile as an appropriate value, but the way the model is constructed using these very conservative values the 95th percentile is no longer than 95th percentile, it might be the 99.9th percentile. Just focussing on the very pages in front of you, at least in this respect reporting data which is in fact at 99.9 per cent as if it were uniformly between 70 and 100 could never be described as conservative reporting, could it? --- Probably not. If this is based on sound data we should actually differentiate between results of simulations and data. They are two different things.' 109 Professor Gardner insisted that the Panel's estimate of a 'moderate' likelihood about R1 and R2 was conservative but indicated he had not formed an opinion about the conservatism of all the assessments made by the Panel; in particular household generation of waste units. 110 Mr Gleeson then took Professor Gardner to the IRAR's discussion of the consequences to small commercial piggeries of an outbreak of PMWS. At p 398, the IRAR assessed as 'high' the likelihood of 'secondary spread (via feral pigs or other means) to a more general population of domestic pigs (including medium - large commercial piggeries)'. Professor Gardner agreed the IRAR did not quantify the effect of such a situation, in terms of either pig deaths or economic losses. His evidence went on: 'What I want to suggest is that if you'd been doing it, Professor, what you would have been more interested in doing, given the limitations of the data, is to form your best judgement, for example, if scenario 4 breaks out, how many pigs are going to die. There's one matter? --- Right, there's a numerical value which would be projections based on the overseas experience. Yes, but you'd like to be able to say, perhaps with a probability distribution, my best estimate is 600 pigs will die and I think the range is probably between 400 and 800. Correct? --- That would be the ideal scenario if there was adequate data to do that. And when you were doing the domestic trade and industry effects, again, you would have liked to say my best estimate is $100 million impact on trade and industry, perhaps varying between 80 and 120. Correct? --- Absolutely, in an ideal world it puts the guess in some measure of uncertainty with it. Well, I just want to ask you about ideal world. Isn't what I've just put to you, namely, trying to put a real measure on these harms, perhaps with a distribution central to a risk analysis? --- Yes. It's usually central to a risk analysis. Again, the issue actually comes to how good the underlying data, and I think that's a questionable issue even in these overseas data with actual quantification of how much PMWS actually happens in these herds and that all goes back, your Honour, to this problem of what actually defines a PMWS pig and what other infections may be present at the same time. What I want to suggest to you is that the failure in this report to make findings on what's likely to occur by way of harm with appropriate probabilities if necessary was a fundamental departure from the essence of risk management. Do you agree? --- No, I don't know that that's fundamentally correct. … Yes. Without having those sort of findings in the report, you just cannot rationally assess the risks arising from allowing this pig meat in. Do you agree? --- I think you can rationally assess it. It's a question of what's been done is put it into a categorisation or box and I think the issue is whether that's necessarily the best choice.' 111 Mr Gleeson referred Professor Gardner back to the 'boxes' in Table 8 (see para 53) and the decision rules. Professor Gardner agreed these rules dictated that, unless there were perceived consequences at a national level, the impact of an outbreak (however severe in a single State) could never be assessed as more than 'minor' (score D). Except in the event, which the professor agreed was unlikely, that the disease affected plant life, the application of rule 8 meant the consequences assessment must be 'low'. When that value fed into the risk estimation matrix (Table 10: see para 39), the result, even with a high likelihood of entry and exposure, was an estimation of 'low risk'. 112 I asked Professor Gardner about the significance of that situation: 'Even if it's 100 per cent certainty that it would happen, by the process that Mr Gleeson was taking you through you have to say it's low and then you've got to say it's a low risk? --- Exactly. So that the X axis drives the final risk estimation value. I think that's a fair assessment. So in a sense all this debate about what is the chance of particular amount of pork being - or meat being contaminated and the chances of feral pigs getting at it and contaminating it, it's all by the way. You [get] the same result by saying any consequences are likely to be less than national level and therefore it's a low risk, end of story? --- Yes, my interpretation would be if it's less than a national level event it's going to most likely be low risk. Well we could take out about three-quarters of the report on that basis, couldn't we? --- It would certainly simplify the real point, your Honour, if that was true.' 113 Professor Gardner told Mr Gleeson that he would not have used a table like Table 10. The professor's evidence went on: Because what's happened under this model is that the consequences of being boxed in before you start in such a way that you're probably not going to get above low, correct? --- If there's not a national outbreak it is unlikely to be above low, that's true. Even if it was 100 per cent chance of entry and exposure and you're told you have to call the risk low? --- The overall risk - the final risk estimate, yes. Even if it was a very high likelihood which it's certainly a reasonably high likelihood, if the consequences are low then the overall final risk estimation value is going to be low risk. Even if you have a low likelihood of entry and exposure which could be up to 30 per cent, which is something you'd normally want to pay heed to, wouldn't you? --- If it was 30 per cent I think that's a reasonably - you know, that's near the threshold with moderate, so if it's right on the cusp it's nearly moderate. But this model tells you that anywhere in low, which is the 5 per cent to 30 per cent, must give you very low risk assuming low consequences? --- Right, but on the other hand if it was - - - Is that correct? --- Yes, that's correct. If it was 31 per cent it would now be low risk. So this model tells you that at 31 per cent you grant the permit and at 29 per cent you don't, is that right? --- If that was the decision rule that would be - and that was a break point right at 30 per cent that would be where you'd actually make the final judgment. Where can you give us examples of other aspects of risk management where 30 per cent is treated as being a relevant barrier at which just below that you run the risk and above that you say too risky? --- I know no similar cases spring to mind where that is a decision threshold that I would necessarily adhere to personally. You have never adhered to that have you? --- Thirty per cent for me is a relatively risky value, but I am risk averse. So just in short you agree that the underlying premise of the model is that for any disease which is going to infect the pig industry, or any other industry that it is applied to, which is not felt in its impact at national level then even though there may be up to a 30 per cent chance of that disease entering and spreading and causing harmful consequences, the model tells you to accept the risk? --- Yes. I mean certainly that is if the consequences are low the only way that you can have a very low or negligible risk is to have a likelihood of entry that is less than low. The model doesn't discriminate between whether it is five per cent or 30 per cent. Wherever it is in that range, in those premises, the model tells you, grant the permit? --- I can't judge about granting the permit, that is a decision rule that someone else is making. But in terms of numerically how this is dealt with my understanding is that that is in fact true; that there is no differentiation. You wouldn't suggest that it was a rational decision which produced that consequence; would you? --- I would have my personal - I would have a different decision threshold to 30 per cent. I know that. I just want to get your answer on this question: you wouldn't suggest that the decision rule which produces the consequences we have just been through is rational; would you? --- I think that is a - my judgment is that that is a threshold that someone else has determined and that is based on their judgement. All I am saying is I might have a different judgment threshold that would be different to the 30 per cent. I might be happier with five per cent as my threshold. You are pretty experienced in the area. Can you point us to some literature anywhere in the world which defends the rationality of that decision rule? --- I can't. In all honesty I can't direct you to any value that says, that is an appropriate level. That is based on the ALOP to my understanding. I can't offer an opinion about whether that has been made on sound basis. Is the answer to my question that you can't point the Court to any such literature? --- No, I can't point it to any such literature.' 114 At the conclusion of Professor Gardner's cross-examination, I put to him some concerns: 'Professor, more than once I think you have used phrases like - I have noted down one time - you said: "Within the constraints of the categories presented." And then you have referred to the constraints of the model more than once. I have an impression that you are really not very happy with the use of this model for the purpose of determining something as significant as the quarantine risk of importing into Australia a hitherto unknown disease. Have I misread you or is that really your position? --- No, I mean I have some concerns about the way it was done. I would have maybe done it slightly differently if I had better numeric values to place on some of these inputs. That is the real problem, there are so many unknowns? --- There are so many unknowns in the problem it makes it very difficult to nail down some of the sub-components. Particularly the R1 and R2 which drive the likelihood part, the L and R calculations. It is very hard to nail it down. Why not put it this way: although you are now resident in the United States you are Australian born? --- Yes I am Australian born. And you know, as we all know, that Australian history is littered with examples of unwise importations into this country. For that reason the PR machine for Quarantine Australia for a number of years has been lecturing everybody on extreme care about importations and that is really almost part of our culture, or at least quarantine authorities would like it to become so; all right? --- Correct. Now, it seems to me, bearing that sort of factor in mind, when you are doing a risk assessment about a product where there is an assessment that there is exposure to a disease which is hitherto unknown in Australia, extreme caution is called for? --- Yes. I mean essentially that goes as the precautionary principle. Right. Well now in this case, rightly or wrongly, the people who assessed this said that they thought there was a high prospect of infected pig meat reaching small commercial piggeries? --- True. Now, is it too simplistic to say, in that situation a person who is doing an assessment for a decision maker and a decision maker would say, well against that background this is an unacceptable risk to take unless I can be satisfied that if indeed the infected pig meat reaches the small commercial piggeries then the consequences are not of any great significance? --- True. I think that is a fair assessment. The difficulty, as you have pointed out, is that because, happily, PMWS is unknown in Australia there is no Australian experience to go on. So the next best thing you can do is to look at the overseas experience? --- Exactly. Wouldn't it have been a more rational way of approaching this exercise for the panel having reached the conclusion that there is a high probability of infected meat reaching Australian pigs, is to say, well let us look at the overseas experience and see what the consequences of that are. Then, having done that, look at the proportion of pigs that have become infected in countries where there is enough data, which apparently includes the United States and I gather Britain, and perhaps Denmark and so on, and it seems to be figures of the order of 20 per cent. Then to work out what the cost is of that, both direct and indirect, in money terms and any other relevant costs. Then at the end of that say, well, if what we think is a high probability if it happens, does in fact happen, this will be the consequences and you have got to weight that against the benefits of receiving this overseas product. Wouldn't that be the approach that you would have preferred to take? --- Yes. I think that is certainly a reasonable approach. The one issue that I think is probably most difficult and hard to get a handle on is the likelihood of this scenario 4 which is a more widespread outbreak. Yes? --- I think the difficulty there is, with the New Zealand experience - if we assume that that is related to pig meat - so far I don't think we have reached a scenario 4 stage in New Zealand, with widespread impact. The overseas experience suggests that most of the outbreaks are related to pig movements and may be other indirect contacts. One of the problems I have about the report is that you can read it looking for that sort of information and you read in vain. It just isn't there. There is no quantification of what this has meant in the countries where there is significant PMWS. What it meant in terms of the impact on individual farms, the industry, the local economy, the national economy. None of that is there? --- Yes. The only material is the material that I have seen under the PMWS block which is page 395 through to - - - Yes and that really doesn't give you the hard facts. It just occurs to me that one would expect a search for information like that and some analysis of it. Particularly in a project that after all was not urgent, it has been going on for some years before they produced the final report. Am I being unfair? --- No. I think that is a fair issue to warrant more detailed consideration. Because I think it is certainly true that the consequences, and where you place the consequences, drives the final value for the risk estimation matrix. That is the primary determinant. 115 I also asked Professor Gardner whether he thought the fact that the cause of PMWS was unknown caused a difficulty in estimating the risk. He replied: 'Fundamentally I think it makes a huge difference in this particular case that if you knew what this additional agent was. Whether it was a new strain of PCV2 or a new unknown virus that would make our job a lot easier because ultimately we would have some diagnostic procedures that could better classify this as PMWS. We would have better data at a herd level. We would have better level data at individual pigs. We might even be able to find out what proportion of pig meat was truly infected at different stages and that would have, I think, a major impact. And you would know what precautions were necessary? --- And what precautions were likely to work, we could do the experiments. So in that sense the fact that we are operating without that information makes this whole problem more difficult. Well what would you say to somebody who said that the proper decision, bearing in mind the quarantine background I have referred to, is to say we don't propose to grant any permits until we know what this factor is. Until then it is a risk we choose not to take and let's do some research by all means. I think it was suggested by Professor Morris there really needs to be a proper study done to try and identify agent X. There's no point in scientists having a squabble about it. They really ought to be finding out what the story is. Wouldn't that be a prudent step? And if this means that there's no decision to grant permits for five years, well so be it? --- I think that's certainly one option. One option would be to assume that it - well there are two options, essentially competing options. One is to say it is a new strain of PCV2 which I think we could logically apply our existing knowledge about PCV2 to that problem. The other is this agent X problem. So some of it actually boils down to how likely you believe one hypothesis is compared with the other. If you believe that it is an exotic strain of PCV2 with probability point nine five then it would seem reasonable to grant permits. If you believed, or there was evidence in favour of this new agent X which we hadn't identified and didn't know anything about and you believe that was probability point nine five, assuming that's my threshold then I would say your decision was correct. So a lot of it actually, in my way of thinking, boils down to which of those competing hypotheses is in fact true.' (c) Dr Gordon Allan 116 Dr Gordon Allan is an honours graduate of the University of Ulster in Northern Ireland. He was awarded the degree of Doctor of Philosophy by The Queens University, Belfast for work on porcine circovirus. In recent times, he has acted as co-ordinator of substantial European Union projects dealing with porcine circovirus diseases. He has visited laboratories in many European countries and acted as a consultant to government instrumentalities in a number of countries. He holds the positions of Principal Scientific Officer for the Department of Agriculture and Rural Development, Northern Ireland and is Honorary Lecturer at The Queens University, Belfast. Dr Allan gave evidence on behalf of the respondent. 117 Dr Allan's evidence was principally directed to the cause of PMWS. He described in some detail the research on this subject that is currently being undertaken, involving expenditure of some millions of Euros. He put that figure into context by citing an estimate 'that PMWS costs around 600 million Euros per year to the European Union', comprising direct losses from mortality of pigs, inability of pigs to reach market weight and indirect losses from increased use of antibiotics and changes in farm practices made in an attempt to reduce PMWS impact. 118 Dr Allan said it was recognised that PCV2 is a necessary factor for the development of PMWS; however it is also recognised that PCV2 must be linked to other co-factors for the full development of the clinical disease. Dr Allan did not offer a firm view about the identity of the co-factors, although he discussed several possibilities. 119 In his briefing instructions, Dr Allan had been asked by the respondent's solicitor to comment upon the risk management conditions recommended by the Panel; in particular, removal of bone, the major peripheral lymph nodes and the head and neck and cooking and curing. In his report, Dr Allan said these measures 'would act to reduce the risk associated with an unknown agent or PCV2'. He commented that pig meat has been imported into Australia from Denmark and Canada for some time, during which those countries have experienced PMWS outbreaks, without PMWS being detected in Australia. 120 Dr Allan gave oral evidence by video-link from Belfast. Much of that evidence concerned his research projects but Mr Gleeson took Dr Allan to his comment about the Panel's recommendations on risk management. Dr Allan agreed the recommended measures would not result in removal of all the lymphatic tissue; viruses relevant to PMWS might remain in the tissue. 121 Mr Gleeson put to Dr Allan that PCV2 may also be found in muscle. Dr Allan responded: 'I mean, I don't know whether it's in muscle or is not in muscle. It may or may not be found in muscle. Nobody has looked.' 122 Dr Allan's evidence went on: 'What I'd like to ask you is whether this is correct, that in all the research that's been done under your EU project, you've never had to scientifically test the extent to which removing the major lymph nodes reduces or eliminates the PCV2 in the balance of the pig meat? --- Correct. And nor have you tested whether, or the extent to which de-boning has that effect. Is that right? --- Correct. Yes. Yes, and you haven't - and because you don't yet know whether there's another infectious agent involved, it follows as night follows day, you haven't been able to test whether removing the major lymph nodes or de boning reduces or eliminates that other virus in the remaining pig meat. Is that right? --- Which other virus are you referring to? Well, agent X, Y or Z. All I'm putting to you is that because you don't yet know what it is, you haven't been able to do that sort of test for it, is that right? --- Well, that's very logical. I mean you can't do tests on things you don't know about. Yes, and because we don't know precisely what X or Y or Z is, we don't know whether it has the same affinity for the lymph node system as PCV2 does, do we? --- Assuming there is an X, Y and Z, no we don't.' 123 Dr Allan made the point that the overwhelming proportion of viruses in pigs, of which he had knowledge, do not replicate in muscle tissue; but he re-affirmed that it is valid 'to assume that if X or Y or Z exists it may also be found in the muscle'. Dr Allan was not aware of any current research on the topic or 'the effect which deboning or removing the major lymph nodes has on the levels of infection in the remaining pig meat'. (d) Professor John Ellis 124 Professor John Ellis also gave evidence on behalf of the respondent. His oral evidence was given by video-link from Saskatoon, Canada. 125 Professor Ellis holds a Chair in the Department of Veterinary Microbiology in the University of Saskatchewan. He was educated in the United States with degrees of Doctor of Veterinary Medicine from the University of Illinois and a Doctor of Philosophy degree, for comparative pathology, awarded by Colorado State University. The professor has been Visiting Scientist at a number of universities and research organisations and has published extensively on topics related to veterinary medicine and animal diseases, including porcine circovirus. 126 Professor Ellis provided a brief, but helpful, history of the growth of knowledge about porcine circovirus and PMWS. He referred to the aetiology and epidemiology of PMWS. He noted that PCV2 is a necessary cause of PMWS. Like other witnesses, he thought it was not a sufficient cause. Professor Ellis said: ' … without the presence of PCV2 in lesions, one cannot make the diagnosis of PMWS, by definition. However PCV2 is apparently not both a "necessary and sufficient" cause, since by itself, PCV2 infection will not result in PMWS. In other words for the syndrome to develop, a co-factor, in addition to PCV2, is required. Examples of co-factors are infectious agents such as other viruses and bacteria, and non-infectious co-factors such as stimulation of the immune system by vaccines, genetics and management practices that are stressful.' 127 Professor Ellis said the impact of PMWS on pig herds is variable. He stated that morbidity was usually 4% to 20%, although it is sometimes as high as 60%. The mortality rate of affected pigs is usually 70-80%, but up to 100%. He said: 'Often, herd performance does not return to pre-PMWS levels. PMWS can have significant economic impact on individual pig herds.' 128 In his report, Professor Ellis said he supported the risk management measures recommended by the Panel. In doing so, he stated his concurrence with the following points, set out in a minute written by Dr David Banks, Chair of the Panel: '-PMWS could be more widespread than reported; under reporting of disease is a relatively common phenomenon in veterinary medicine, especially when there are costs involved in obtaining specific diagnoses in animals that have finite (market) value. -pig meat could be subclinically or persistently infected with heterologous isolates of PCV2 (different from those in already in Australia) or with other agents that could serve as co-factors in the pathogenesis of PMWS. -PCV2 or unknown agents could certainly be present in meat, particularly in lymphoid tissue. -PCV2 is a very stable virus that could persist in carcasses. Moreover other viruses, for example parvoviruses, that are recognized co-factors in the pathogenesis of PMWS could also be stable in carcasses. -Waste meat discarded could initiate infection if eaten by a pig and there are classic examples that demonstrate that phenomenon, including, foot and mouth disease and African swine fever. Since there are likely to be other agents than those already identified that could play a role as co-factors in the pathogenesis of PMWS, the worst case scenario adopted by the IRA are justifiable. -Because some identified infectious co-factors, as well as those still to be identified could remain viable prior to scavenging by a feral pig, the worst case scenario adopted by the IRA are justifiable. Again there is biological precedent that documents that latter phenomenon in the initiation of epidemics of swine disease.' 129 In cross-examination, Mr Gleeson discussed with Professor Ellis a number of possible co-factors of PMWS, some being infectious agents and some non-infectious. Professor Ellis agreed with a suggestion 'that it is possible that there is an as yet unidentified infectious agent which may in some circumstances be the trigger to produce PMWS'. The professor said there could be more than one unidentified infectious agent, possibly different infectious agents in different parts of the world. 130 Mr Gleeson asked Professor Ellis whether he had any idea how many herds in Canada were affected by PMWS. The professor said he did not really know; there is virtually no data. However he said he would guess 'it might be upwards of 75 per cent of herds, it might have had one or a small number of cases and a much smaller percentage that it would have had serious problems with'. Professor Ellis thought he recalled seeing data from Iowa, in the United States, showing that 25 to 30% of herds are PMWS infected. 131 Mr Gleeson questioned Professor Ellis about the adequacy of the risk management conditions recommended by the Panel. The professor agreed that removal of the major lymph nodes from a pig would not entirely eliminate the lymph system from the body. Mr Gleeson suggested that 'even if you tried to get every lymph node out of a pig, you'd probably only get about 70 or 80 per cent'. Professor Ellis responded that it 'depends on the cut'. The evidence went on: 'Yes, and when we are concerned with viruses within say pig muscle, and their potential for transmission, you don't need to have a great deal of the virus in order for transmission to occur, do you? --- That's correct. So that it's not as if you need to have the virus multiplying within the pig meat. It's enough if it's there and it remains alive in a small dose, capable of being transmitted. Correct? --- In theory, yes. … I'm sorry, I take it there are logically three possibilities, a brand new virus group at one end. Secondly, a variant on a known group and at the other end a known but unrecognised virus. Do you agree with that? --- Correct. Now, if I can just contemplate all those within the questions I'm asking, I'm suggesting that that unknown virus which I will call it may well be found in the muscle of the pig which has PMWS? --- That could be the case. Yes. Yes, and if it's in the muscle then if that muscle travels through the waste system and gets before a healthy pig, that could cause the transmission of the virus. Correct? --- Yes. Yes, it could. Now, the fact that you have removed the major lymph nodes before you got to the pig muscle doesn't allow you to conclude that you have removed the unknown virus from the pig meat, does it? --- Removed completely, probably not but removed substantially, probably yes - by removing lymphoid tissue and the blood.' 132 Professor Ellis later agreed with a proposition put by Mr Gleeson that: 'since we don't know what the unknown is … we can't safely conclude that the level of the unknown in the muscle is no greater than the level of the PCV2 virus in the muscle'. 133 Professor Ellis agreed the level of the unknown agent in the muscle is a different thing from its hardiness. Both matters were referred to in this exchange: 'You are of the view, aren't you that PCV2 is a very hardy virus? --- Yes, it's probably one of the most hardy viruses that we know of. And by that we mean that it's resistant to environmental degradation? --- Correct. That covers the concept that if we got PCV2 in pig meat there is a good chance that as pig meat travels through the system of consumption and waste the virus will still be there at the end of the day to do its work, correct? --- Yes. Again, those kinds of experiments really haven't been done but certainly I would predict that PCV2 would, again, it stands a lot of environmental attempts at degradation. One of the things you agree with the IRA report on is this, that because PCV[2] is so hardy if there is an unknown virus involved it would be reasonable to think that it will not be more hardy than PCV2, correct? --- Yes, I would say the odds that it would be more hardy than PCV2 are low, I would say. Although you've expressed that opinion supportive of the report I would suggest that you do not agree with reasoning that suggests that the amount of the unknown virus in the pig muscle can be assumed to be the same as the amount of PCV2? --- Yes. Again, it would depend on the particular type of virus that it is what cells it lives in, that would be the primary determining factor probably of how much there would be in the meat. You would agree we just don't know about that at the moment, correct? --- That's correct. Finally, if you were contemplating measures to reduce the chances that pig meat coming into Australia has infection in it derived from the PMWS syndrome, when you are looking at the pathways, one of the critical questions is how likely is it that there will be enough of the relevant virus in the pig meat to cause a problem when it travels through the waste system, correct? --- Correct, that's what we are really talking about at the end of the day, what are the odds of that happening. We're obviously in a difficult area because we have to hypothesise that we may be dealing with a virus as yet unknown, you agree with that? --- Yes.' 134 It is appropriate to note Professor Ellis' answer to Mr Gleeson's final question (allowed over objection) on this topic: 'I will just ask you finally, doctor, if you would agree with this proposition? Because it is reasonable to contemplate that there might be an unknown agent involved with PMWS and because of our lack of clear knowledge on what the agent is, how it operates, and how much of it may be found in the muscle we would have to be pretty conservative when we were forming views on whether risk measures were likely to be effective. … Well the specific answer, the yes/no answer to your question is, yes. But again I think that the measures that have been designed and applied here are conservative and in science and in life it is about playing the odds, right. So, again, if you take the lymph nodes away, you take the blood away, the odds are low that you are going to have enough of an infectious agent of any type to be a problem.' The professor explained that the phrase 'the odds are low' meant the chances are low. (e) Mr David Vose 135 Mr David Vose has degrees of Bachelor of Science in Physics and Master of Science in Physical Oceanography from two English universities, Durham and Southampton respectively. After some years of employment in New Zealand as a risk analyst, in 1993 he set up his own risk analysis company, Risk Media Ltd. That company has carried out risk modelling for a wide range of corporations and some governments. Mr Vose has also conducted risk analysis training seminars and provided expert guidance to some international agencies. Mr Vose gave evidence on behalf of the respondent. 136 In his report, Mr Vose expressed the opinion that the methodology employed by the IRAR: 'does indeed provide a sound and rational method for: representing, reporting and taking account of uncertainty in expert judgments; assessing the quarantine risks associated with the hazards under consideration; and reaching conclusions about whether those risks were "acceptably low" (i) in the absence of any risk management measures; and (ii) as modified by identified risk management measures.' Mr Vose went on to state his reasons for that opinion. 137 The first matter discussed by Mr Vose was the IRAR's reliance only on the 50th percentile as opposed to the 95th percentile. Mr Vose defined the terms. He described the 50th percentile as: '[a] measure of the mid-point of the calculated distribution that one believes the true likelihood has equal chance of lying above or below'. He said the 95th percentile is '[t]he value that one believes the quantity being modelled has a 95% probability of being less than'. 138 Mr Vose noted Professor Pettitt's criticism of the IRAR's reliance on the 50th percentile and commented: 'The choice of selecting the 50th percentile rather than the 95th percentile of the distribution of the likelihood estimate is a reflection of Biosecurity Australia's risk attitude (i.e. how it values risk). If Biosecurity Australia were to change its selection from the 50th percentile to the 95th percentile, it would move many if not all of the estimated disease risks up one or more categories (for example from "Very low risk" to "Low risk" or "Moderate risk"). If Biosecurity Australia wished to maintain the same level of biosecurity as it maintains now (and if it wished to maintain consistency with other decisions it may already have made) it would be necessary to increase its appropriate level of protection (usually abbreviated to "ALOP") from "Very low risk" to a higher category (e.g. "Low risk") to compensate for reading the 95th instead of the 50th percentile.' (Footnotes omitted) 139 Mr Vose stated it is 'entirely within Australia's sovereign rights to select the percentile that it deems most appropriately reflects its attitude to risk. The selection [of] the reference percentile forms part of the ALOP'. He later said: 'There is no right or wrong to the selection of the percentile to use; it is a matter of choice, reflecting one's risk attitude'. 140 Mr Vose concluded on this issue: 'From the point of view of a scientifically-based analysis, the choice of the 50th percentile underlines that the analysis is attempting to take a neutral position, evaluating the likelihood of entry and exposure at a point where it is estimated that there is a 50% chance the true likelihood is larger, but also a 50% chance the likelihood is smaller. From a mathematical point of view, the 50th percentile and the mean (the x-axis "balance point" of the distribution) are two stable statistics that provide a good representation of the central location of a distribution. The key advantage to using the 50th percentile over the mean is its consistency of interpretation: the mean does not correspond to anything probabilistic that is intuitively easy to interpret, which makes its use vulnerable to being misunderstood. One argument put forward by Biosecurity Australia for using the 50th percentile is that it is a stable statistic, meaning that it takes relatively few iterations in a Monte Carlo simulation for this statistic to reach its final value. However, I ran the model on my laptop PC and it took just 35 seconds to run 10,000 iterations, which is sufficient to get a fairly precise 95th percentile. So, like Prof. Pettitt … I think that the stability of the statistic is not an important factor, although the points I mention above provide a substantial rationale for selecting the 50th percentile. Summary: The 50th percentile is a statistic that is intuitively easy to understand, which means that more people will be able to follow the import risk analysis methodology. It is also a reflection of the risk attitude of Biosecurity Australia, and must be viewed not in isolation, but together with how Biosecurity Australia addresses uncertainty in its estimates, and the level of risk ("Very low") that it deems acceptable, since all these components contribute to describing Australia's ALOP.' 141 The second issue addressed in Mr Vose's report was whether the methodology employed in the IRAR 'was capable of providing a reasonable assessment of the level of quarantine risk associated with the importation of pig meat into Australia, bearing in mind Biosecurity Australia's aim of ensuring a consistent approach to decision-making across all IRAs undertaken by it'. 142 Mr Vose responded: 'The methodology provided in the Guidelines offers reasonable flexibility to quantify unrestricted and restricted risk. It advocates risk analysis techniques that are the accepted norm. The IRA is based on sound mathematics, and if all other risk assessments based on the same methodology maintain the same mathematical quality, I believe that the methodology will provide a reasonable assessment of quarantine risk. The Guidelines introduce the idea of a generic risk assessment that establishes sanitary requirements that would protect Australia's ALOP no matter which country was exporting to Australia. This should open up trading opportunities for any exporter that is willing to accept the standard sanitary measures, and minimises the bureaucracy and delay in receiving import approval. Moreover, it has devised a logical and transparent method for relating an estimated risk to the Australia ALOP. This method can be used as a very effective tool to demonstrate consistency in decision-making across all animal and animal product imports, all pathogens and for all countries.' 143 Mr Vose next turned to the complaint that the IRAR assessed the risk of exposure on an annual basis. He claimed annual estimation was common. He said: 'Reporting a likelihood over a longer time period (a century, for example) provides no greater accuracy to the estimate, and may perhaps give the reader a false impression of the stability of the risk which, by the nature of the development of new pathogen strains, the possibility of new sanitary measures, and free trade is inappropriate. In international trade, for which "per year" is the norm, I think that estimating the probability for longer periods would be seen as attempting to deliberately exaggerate the risk level, and thus undermine the need for Biosecurity Australia to remain, and be seen to remain, disinterested.' 144 Mr Vose referred to a submission that APL had made in relation to a draft of the IRAR; that use of an annual risk estimate underestimates the risk. He said: 'The example is given that a likelihood of entry and exposure of 0.027 would be considered "very low" (which they comment "sounds disarmingly reassuring"), but if one looks at the likelihood over 10 years it becomes about 0.27, which they consider should be characterised as "low" and after 50 years they consider would be "high". Obviously the probability of introduction increases the longer the duration of exposure, so no matter the level of risk one can arrive at a probability near 1 if one considers a sufficiently long period. The annualised risk helps avoid selecting arbitrary durations, and allows a comparison between risks. The risk would not change category, however, by increasing the duration over which one calculates the likelihood of introduction and spread because this likelihood is a rate, i.e. it has a denominator which is the number of years of exposure, which always brings it back to the same value.' 145 In summary, Mr Vose said: 'For the reasons given above, I believe that Biosecurity Australia was correct in using an annual estimate of the likelihood of entry and exposure. It has used qualitative judgement of the consequence of disease entry because of the inability to do this quantitatively, and taken into account the possible duration of outbreak consequences. I therefore believe that the Biosecurity Australia methodology has chosen the most appropriate time period for its analysis. It would be no more meaningful to provide an estimate of the risk over 10, 25 or 50 years than on an annual basis.' 146 The next section of Mr Vose's report discussed whether the IRAR's methodology was consistent with guidelines issued by certain international organisations. Neither party has agitated that issue. 147 Mr Vose gave oral evidence by video-link from Italy. Before he was cross-examined by Mr Leeming, I asked him to explain the Monte Carlo risk system. He replied: 'Well, Monte Carlo simulation is a term that was coined way back in the Manhattan Project for the development of the atomic bomb. It was given the name Monte Carlo because of the Monte Carlo city which is well-known for its casino gambling but the idea is to randomly generate numbers with a computer to try to estimate probability distributions of the outputs you're interested in. The reason you use Monte Carlo simulation is because to do it mathematically with algebra it becomes mathematically intractable, impossible to do, and you can bypass the difficulty of the mathematics using random number generation to create distribution.' 148 Mr Vose confirmed that the idea was to keep running the computer while it randomly selected values within the selected range; for example, between 30 and 70 for 'moderate'. He explained: '… normally, you would have several probability distributions in the same model and all of the distributions would be sampled randomly at the same time to calculate an output then by running many iterations, many, many different scenarios, in other words lots of random numbers, you create an output distribution which is simply normally just a histogram of those generated values. The 50th percentile is the value of which 50 per cent of those generated output values lay below and, therefore, 50 per cent lay above. The 95th percentile would be the value at which 95 per cent of your iterated value fell below and 5 per cent fell above. 149 Mr Leeming took Mr Vose to the comment in his report (para 138 above) about the consequence of Biosecurity Australia changing from the 50th percentile to the 95th percentile. He said this would not affect the inputs, which must be estimated without regard to the final answer, but it would involve moving the probability bands in terms of what is considered tolerable. In other words, as I understand Mr Vose, a decision-maker who would accept only a low risk on a 50th percentile basis might accept a moderate risk on 95th percentile. 150 Mr Vose made it clear that the selection of the appropriate percentile was not a matter for the risk analyst; it was something for the risk manager to determine, in deciding what degree of risk was acceptable. Mr Vose agreed that the reference in his report to the IRAR having adopted conservative input estimates was not something within his own expertise or knowledge; he had relied primarily on claims of conservatism within the IRAR itself. Mr Vose noted one aspect which he thought to be mathematically conservative. However, he agreed with Mr Leeming that, if some estimates were conservative and some were not, it was not possible to have the same degree of confidence in the median. 151 Mr Leeming suggested to Mr Vose that not all the estimates were conservative. For example, item R5 in the release assessment assessed as 'high' the likelihood that the pathogenic agent will not be destroyed by the post-mortem decrease in muscle pH that accompanies carcass maturation. Yet, Mr Leeming suggested, this was almost a certainty, given the facts (stated in the IRAR) that PCV is stable at pH3 and meat is not assumed to reach a pH lower than 6.2. Mr Vose agreed that, on the basis of these statements, R5 is a 'pretty certain outcome'. Mr Vose made a similar concession in relation to R6 in the release assessment, dealing with the likelihood that the pathogenic agent will not be destroyed during cold storage and transport. 152 After consideration of the spreadsheet showing the Monte Carlo distribution of the likelihood of events, Mr Vose agreed that an adjustment from 85% (the mid-point of the 'high' range) to 99% in relation to each of R5 and R6 would increase the mean points of the calculation (0.08 on the 50 percentile and 0.16 on the 95th percentile) by almost 30%. 153 Mr Leeming suggested to Mr Vose that the IRAR could best be described as 'semi-quantitative'. Mr Vose said he would 'prefer to use my own terms and say it was a blend of quantitative and qualitative but, yes I'll call it semi-quantitative if you like'. Mr Vose affirmed, as still true, a statement he had made in a paper published in 2001 that 'semi-quantitative risk assessment techniques are commonly used for risk analysis in commercial projects but are currently not widely accepted in international risk issues'. 154 When asked about the appropriateness of estimating risk on an annual basis, Mr Vose spoke of financial decisions being made by reference to net present value; that is, the value of particular items in the current financial year. 155 Finally, Mr Leeming referred to a comment Mr Vose had made, in response to Professor Pettitt, about the desirability of 'using some sensitivity analysis on the likelihood category: for example where "negligible likelihood" is being used, replace the estimate with "extremely low" and determine whether the risk management decision remains unchanged'. Mr Leeming obtained Mr Vose's affirmation of a statement Mr Vose had made in a book: 'The uniform distribution is generally a very poor modeller of expert opinion since all values within its range have equal probability density but that density falls sharply to zero at the minimum and maximum in an unnatural way.' 156 The evidence went on: 'And one of the vices of a semi quantitative analysis is that when you have to put an outcome of the Monte Carlo simulation into a box and the box is defined precisely say between point three and point seven, or point seven and point one, then I think as you said a few times in answer to me, you just follow the rules and the rules determine what the outcome could be, and the reason that you are nodding with me as I ask you that question is that that is unnatural and doesn't have regard to the fact that 71 percent for example is right on the cusp between, in this model, moderate and high verbal descriptions? --- And I think you will find that if you use any probability distribution there will be a value that is on the cusp. But the position is particularly extreme, isn't it, with uniform distributions because as you and I both know when one is combining probability distributions in models such as this a uniform distribution is a most unlikely outcome to fit a curve into, isn't it? --- Yes, and if you read some other parts of my book there, you will see me saying that one use, perhaps the only really good use of a uniform distribution is when you want to exaggerate your uncertainties and it allows you to do a sensitivity analysis.' 157 At the end of Mr Leeming's cross-examination, I asked Mr Vose about the use of a one-year estimate of risk in this type of case. He said he thought it 'very inappropriate to try to bring … the consequence analysis down to numbers'. He described the method of calculating the chance of an event occurring over ten years: if there is a 30% chance of something happening in one year, there is a 70% chance of it not happening. To determine the chance of the event not happening over ten years it is necessary to multiply 70% ten times. This yields the result that there is only about a one in 32 chance that the event will not happen. Another way of stating this is to say there is a 31 out of 32 chance (96%) that the event will happen at least once in ten years. [A comparable calculation in relation to the mid-point of the 'low' range used in the IRAR i.e. 0.05 to 0.3 shows that, over ten years, there is only a 16.2% chance that the event will not happen. In other words, there are five chances out of six that the event will happen.] 158 After some discussion between Mr Vose and myself about the width of the ranges used in the IRAR, especially for 'moderate' risk, Mr Leeming suggested to Mr Vose that the IRAR was different from any other risk analysis he had seen internationally. Mr Vose responded: 'Yes, because no other risk analysis I've seen has had the guts, if you like, to address the consequence component of the risk assessment. They all seem to stop, if they even go this far, they stop at quantifying the probability of a disease introduction and just say well that's above or below an acceptable probability, which I'm afraid to say doesn't match OIE guidelines that we all subscribe to. But it's not merely the consequence analysis, also what makes this analysis different from others on the international stage is the semi-quantitative analysis at the front end of the model, that's something you haven't seen before? --- Well, I haven't seen in the import risk analysis arena I've seen it in other arenas, yes, but not in this arena. You may have seen it in industrial applications, for example commercial industry, but not in risk analysis? --- Not in animal import risk analysis. Yes? --- I've seen it in projects and commercial risk analysis yes, not in animal import risk yes, no.' (f) Professor Morris in reply 159 Professor Morris provided a report in reply to the reports of the respondent's experts. Much of this report concerned the likely cause of PMWS. It is not necessary to refer to that material. However, it is desirable to mention some other points. 160 Professor Morris noted that some of the respondent's witnesses argued it can be inferred there is only a low risk of importation of PMWS infected pig meat from the fact that Australia has imported pig meat for several years from countries in which PMWS has been diagnosed, but with no subsequent outbreak in Australia. He said: 'This is flawed in various respects. Firstly, I have examined the import statistics and the amount of imported product has only recently begun to rise steeply from earlier low levels. Figure 4 on page 37 of the IRA demonstrates this in crude summary form. Secondly countries such as Denmark have only relatively recently become infected with PMWS, so there has been a significant increase in the release likelihood. Therefore past history does not give a good guide to future risk.' 161 Figure 4 in the IRAR shows Australian pig meat imports running at about 10,000 tonnes per annum in January 1998, rising to about 30,000 tonnes in January 2001 but exceeding 50,000 tonnes by the end of 2003. 162 As a further response to this approach, Professor Morris attempted to estimate the probability of a scenario 4 outbreak (secondary spread to a more general population of domestic pigs, including medium-large piggeries) using the values adopted in the IRAR, and making the uncontroversial assumption that the average weight of a deboned carcass is about 50kg. Professor Morris found that, considered over a five-year period, the IRAR's estimate of a 'low' risk (before application of the conditions) becomes 'moderate' at the 50th percentile and 'high' at both the 75th and 95th percentiles. He said: 'It must therefore be concluded that the proposed risk management strategy is unlikely to prevent entry of PMWS for more than a few years, given the expected volume of trade and the occurrence of PMWS in source countries. The analysis also makes the assumption that the agent of PMWS is largely limited to lymphoid tissue. If the agent can also be found in muscle tissue, the basis of the risk management strategy is seriously undermined.' This conclusion was not challenged by counsel for the respondent. (iii) Circumstances surrounding the Permit Decision 163 As mentioned in para 3 above, the Director determined on 10 May 2004 that he would adopt recommendations contained in the IRAR as a 'policy … to be taken into account by decisions makers' in accordance with the Act and Proclamation. 164 The delegate who issued the Fayman permit was Andy Carroll, an officer of the Australian Quarantine and Inspection Service ('AQIS'). The documentary evidence shows that, on 27 July 2004, Dr Carroll was provided with a minute from another AQIS officer, Patricia Thornhill recommending that he grant the permit. Ms Thornhill referred to the release of the IRAR, the dismissal of six appeals by the Import Risk Analysis Appeals Panel challenging the recommendations in the IRAR and the IRA Decision of 10 May 2004. She set out certain information received from the proposed exporter country, the United States. 165 After referring to the relevant statutory provisions, Ms Thornhill advised Dr Carroll: 'You need to be satisfied that the imposition of the conditions set out in Attachment B on the import permits will limit the quarantine risk to one that is acceptably low and that you are not aware of any reason not to adopt conditions derived from the Final Report [that is, the IRAR].' 166 Ms Thornhill attached to her minute a number of documents. They included, as attachment G, what Ms Thornhill described as an opinion 'by counsel briefed by the Department in the matter' concerning the merits of the present proceedings. Because of their desire to maintain their clients' legal professional privilege, counsel for the respondent did not include the opinion in the tendered documents. However, they did include a memorandum from Dr Carroll to Dr Banks in which he said: 'I have read the submission of 27 July 2004 from Ms Patricia Thornhill recommending that I exercise my delegation under Section 13 of the Quarantine Act 1908 to grant Import permits for pig meat from the USA. I have considered the opinion at Attachment G to the submission from counsel briefed by the Department in relation to the challenge by Australian Pork Limited and others in the Federal Court of Australia. I have concluded that I require confirmation that the Import Risk Analysis Panel did approach their consideration of the issues along the lines set out in paragraphs 2.5-2.8 of that opinion. I would appreciate your advice on this matter.' 167 Apparently as a result of this memorandum, Dr Banks convened a teleconference of the Panel on the following day, 28 July. On 29 July 2004, Dr Banks sent a memorandum to Dr Carroll. He stated the purpose of the teleconference had been: 'to discuss whether any new significant information had become available on the aetiology of PMWS and to set out the consideration that the Panel had given to the hypothesis that an unknown disease agent/s may be the trigger for activation of PCV2 and expression of clinical PMWS in the final IRA of pig meat.' 168 Under the heading 'Considerations', Dr Banks said: 'Biosecurity Australia is unaware of any new significant findings on the aetiology of PMWS since the finalisation of the IRA report of pig meat. Scientific information will continue to be monitored. In conducting the risk assessment either different virulent strains of porcine circovirus type 2 (PCV2) to those present in Australia or an unknown disease agent/s in conjunction with PCV2 were considered by the risk analysis panel (the Panel) as the cause of PMWS and assessed. It was considered that if an unknown disease agent is involved, it is most likely a virus and is not present in Australia. Information on PMWS was assessed to determine the likelihoods assigned to each step in the release and exposure assessments. The import risk analysis also examined the consequences of establishment or spread of PMWS. Where it was necessary to consider agent information in the assessments this related to PCV2 as it was considered that this would also encompass an unknown disease agent. Due to the characteristics of PCV2 and subclinical and persistent PCV2 infection the IRA adopted a worst-case scenario compatible with the known scientific information. Throughout the PMWS risk assessment conservative estimates were made as follows: • PMWS could be more widespread than reported; • Pigs could be subclinically and persistently infected with either virulent PCV2 or an unknown disease agent; • Virulent PCV2 or an unknown disease agent may be present in meat or associated tissues, in particular lymphoid tissues as viruses have an affinity with these tissues; • PCV2 or an unknown disease agent is a hardy virus and could persist in carcasses during maturation and cold storage; • Meat waste discarded could initiate infection if eater by a pig; and • PCV2 or an unknown disease agent is a hardy virus and would remain viable prior to scavenging by a feral pig or swill feeding of backyard or small commercial pigs. The Panel considered the risk management measures below, which addressed virulent PCV2, would be equally applicable to an unknown disease agent in PMWS and would act to reduce the risk associated with that agent: • Removal of lymphoid tissues (risk tissues - bone and major peripheral lymph nodes) - many disease agents have an affinity for these tissues and hence the recommendation for their removal; • Commercial cooking or long term curing - most disease agents are inactivated or partially inactivated at commercial cooking temperatures or during long term curing; • Removal of risk tissues and cooking or curing reduce the volume of waste discarded and hence exposure opportunity.' 169 On the following day, 30 July 2004, Dr Carroll granted to Fayman a permit to import into Australia from the United States an unspecified quantity of uncooked, boneless pig meat for human consumption. The applicants' submissions 170 Counsel for the applicants broke their submission into four parts. First, they criticised the manner in which the IRAR translated its unrestricted risk assessment into its conclusion about restricted risk. Counsel contended the IRAR dealt with the critical issue of the effect of the restrictive conditions by reference to PCV2, rather than to PMWS, and failed to consider the restrictions necessary to guard against the importation of PMWS into Australia through imported pig meat. As a result, they said, the IRA Decision involved error of law, failure to take into account a relevant consideration and a decision that was not justified by the evidence. 171 Second, counsel argued the IRAR's assessment of unrestricted risk was defective for the following reasons: it assessed only annual risk, not longer term risk; it substituted computer model based questions for the statutory question; and it failed to consider what benefits (if any) were to be balanced against the accepted risk of importing PMWS affected meat. Counsel said each of these matters constitutes a failure to take into account a relevant circumstance. Moreover, the IRA Decision was defective because the statutory question, under s 70 of the Act, had to be answered in respect of each individual permit. As the incidence of PMWS varied from country to country, this meant there could be no general rule; each application needed to be considered on its individual merits. 172 Third, the IRA Decision was an improper exercise of power because it was so unreasonable that no reasonable person could have reached that decision, having regard to the defects of the IRAR. 173 Finally, counsel put submissions concerning reviewability of the IRA Decision. It is common ground between the parties that the Court has power to review the Permit Decision. 174 I will deal separately with each of these parts of the submission: (i) The IRAR consideration of restricted risk 175 Counsel commenced this part of their closing submission by emphasising four non-controversial points: (i) Australia is currently free of PMWS; (ii) PCV2 is a necessary, but not sufficient, factor for the expression of PMWS; (iii) PCV2 is already present in Australia; and (iv) the additional factor (or factors) necessary for the expression of PMWS has not yet been identified. Among the possibilities are an infectious agent or agents (a DNA virus, an RNA virus or a bacterium), a non-infectious agent or agents or any combination of these. 176 Counsel said that, against this background: 'a critical question for the IRA Report, when considering the risk measures, was to identify and evaluate measures which would reduce the quarantine risk for the import of pig meat which may come from a pig affected by PMWS to a level which was acceptably low. The Report could not simply assume that the risk arising from import of pig meat from a PMWS infected pig was the same as a PCV2 infected pig. If it was, the whole exercise was unnecessary.' 177 Counsel referred to Professor Gardner's comment, quoted in para 115 above, that 'a lot of it' [that is, choosing the prudent step] 'boils down to which of those competing hypotheses [as to the cause of PMWS] is in fact true'. 178 Recognising the practical problem confronting the Panel, because of uncertainty about the competing hypotheses, counsel drew attention to material in the IRAR concerning PRRS. The IRAR records that Biosecurity Australia commissioned a Netherlands company to conduct an experiment in that country in which 12 pigs were inoculated with PRRS virus and slaughtered eleven days later. PRRS virus was detected in the semimembranosus muscle of seven of the pigs. Infected muscle from those pigs was then fed to healthy pigs. Collected sera demonstrated that strains of PRRS virus had been transmitted to the receiver pigs. The Panel used the results of this study in determining the import conditions necessary to guard against the PRRS virus entering Australia. 179 Without presuming to design the experiment, counsel suggested it might have been useful for the Panel to arrange a similar study of PMWS. Unlike the case of PRRS, the infectious agent for PMWS (if any), other than PCV2, is not known. However, counsel suggested, it would have been possible to feed a group of healthy pigs PMWS infected meat that had been treated in various ways, including by cooking to various temperatures. Dr Allan agreed it would be 'relatively simple' to design a project to determine whether feeding healthy pigs meat taken from PMWS-affected pigs led to transmission of the syndrome. He also agreed it would be possible to extend the project by feeding selected pigs PMWS affected meat heated to various temperatures. 180 This submission led to a central point in the applicants' case: the Panel's lack of information as to the restrictions necessary to guard against importation of PMWS into Australia. Counsel pointed out there is no information in the IRAR as to the level of PCV2 in the muscle of a pig after removal of its lymph nodes. At p 389 of the IRAR, the Panel said it was 'unaware of any studies that have examined skeletal muscle for the presence of PCV2 viral antigen or virus'. Neither Professor Ellis nor Dr Allan knew of any such study. Nor was there any information about the quantity of PCV2 necessary to infect consuming pigs. Counsel submitted: 'If there is no information on the level of PCV2 virus which is necessarily present in muscle in order to transmit infection, a fortiori there is no basis to conclude what level of PCV2a or agent X would need to be present to allow transmission.' 181 Accordingly, counsel argued, 'there is no basis to conclude that L2 would be reduced from moderate to very low' by the recommended restrictions. 182 Counsel spent some time examining the risk management section of the IRAR. They noted that, at p 746, the IRAR stated the direct effect of processing on PCV2 survival had not been examined. They said: 'Thus, unlike the case of PRRS where a study was done to see whether cooking to a certain temperature for a certain period would inactivate the virus …, no such study was done for PCV2. A fortiori, no study was done, and there was no basis to conclude, what affect cooking would have on PCV2a or agent X.' 183 Counsel then referred to a statement made at p 747 of the IRAR: 'As PCV2 has been shown to result in persistent infections, has been isolated from many tissues, is a hardy virus and is likely to be transmitted orally, the Panel considered that if an unknown disease agent was involved in PMWS, the risk management measures requiring removal of bone, major peripheral lymph nodes, head and neck and cooking or curing would act to reduce the risks associated with that agent.' 184 Counsel said: 'The reasoning expressed is that because of certain features of PCV2 (in particular its hardiness as a virus), the proposed measures would reduce the risks associated with any agent X. However, if, as is the case, no scientific studies have been done on the level of PCV2 which needs to be present in muscle to transmit the virus, and a fortiori no studies have been done on PCV2a or agent X on this topic, there is no basis to conclude that the so-called "reduction" of the risks would be of the same level as might occur with PCV2. Nor does the report clearly do so. The question must be asked: why did not the panel commission a study requiring the cooking of PMWS meat to determine at what temperature and over period of time would transmission be prevented? Had this been done, the panel would have directly considered the effectiveness of the proposed risk measures against the very subject at issue, namely PMWS affected meat, rather than making inferences as to whether the position with PCV2 - itself not scientifically demonstrated - might carry over to PMWS. Such a project would be relatively simple.' (Original emphasis) 185 Counsel noted that a project such as this is contemplated in the SPS Agreement, referred to at para 36 above. The feasibility of such a study was supported by both Professor Morris and Dr Allan. There was no contrary evidence and no explanation as to why such an experiment was not required or commissioned by the Panel. 186 Counsel submitted these matters raised grounds for judicial review of the IRA Decision: '(a) the error of law is that at this critical point in the process, namely attempting to determine the level of quarantine risk with the measures in place in respect to meat which may be PMWS affected, the Report has reasoned on the basis of what the position might be with PCV2 rather than focussing squarely on the identification of, and analysis of the transmission of, the infectious agents which may be present in the PMWS affected meat. The critical step in the reasoning is suspect so far as PCV2 is concerned because of the lack of scientific knowledge as to the infectious dose of PCV2 likely to be found in muscle after the measures. No conclusions can be drawn from that insecure foundation as to the likelihood of an infectious dose of the unknown and unidentified PMWS agent remaining in pig meat after the measures, especially when the report has not addressed the probabilities of what the infectious agent is; (b) alternatively, this is a failure to take into account a relevant consideration, namely whether the measures imposed would address the risk of the infectious agent which underlies PMWS affected meat; (c) alternatively again, it is a decision made without evidence on this critical point.' (ii) Assessment of unrestricted risk (a) Annual basis 187 The Panel considered the risk of importation of PMWS only on an annual basis. Counsel for the applicants submitted this limitation was unjustified. They argued that, although s 70 of the Proclamation requires the level of quarantine risk to be considered in relation to each permit: 'it clearly permits, and indeed requires, that consideration to take into account the practical realities that the risk is affected by other permits granted, or likely to be granted, over a relevant period applying the same conditions. There is nothing in section 70 which specified that one year is the maximum period over which the risk must be assessed.' 188 Counsel submitted that a relevant consideration for the decision-maker, in regard to an application for any permit, 'is what the risk would be if a longer period were considered, and the significance to be attached to that additional risk'. They acknowledged that Professor Pettitt conceded that taking a longer period of time increased uncertainty, but they noted he also stated, in para 5.7.6 of his report, that 'as the volume of imported pig meat increases so does the number of waste units, and so does the likelihood of there being one or more outbreaks, and so does the overall risk'. 189 Professor Pettitt noted that Table C, at page 145 of volume 3 (Annexes) of the IRAR, set out the likelihood of one or more outbreaks over ten years as being 99% at the 50th percentile (100% at the 95th percentile) compared with only 38% at the 50th percentile (95% at the 95th percentile) over one year. These assessed likelihoods all assume adoption of the Panel's recommended conditions. The Table assumes no variation in the annual volume of trade over the ten year period; if volume were to increase, the risk would be even higher. 190 Counsel referred to the Annexes volume of the IRAR which contains responses by the Panel to some submissions made to it in relation to a draft of the IRAR. At p 138, the Panel responded to an APL submission criticising consideration only of annual risk. The Panel said: 'The basic tenet of the comment is that, all things being equal, risk increases with the volume of product imported. As the volume imported increases, the likelihood of pest or disease introduction gets closer to one. Australia has a managed risk policy for biosecurity risks, it is not a zero risk based policy. The ALOP is based on annual risk, thus it is appropriate to compare the calculated annual risk to the ALOP.' Counsel commented: 'By definition, nothing further will be considered.' 191 Counsel also noted the final sentence in the Panel's response: 'It should be noted that pig meat has been imported into Australia for 13 years from countries where both PRRS and PMWS occur, with no exotic disease outbreak occurring.' 192 They said: 'This conceals the fact that the disease [PMWS] has only been identified in Canada from 1996, Denmark from 2000 and New Zealand from 2003, and there has been a surge in volume of imports in the last 2 years.' 193 Counsel for the applicants also noted Professor Morris' evidence in reply about the likelihood of a scenario 4 outbreak of PMWS over the next five years. This calculation adopted the Panel's assumptions about volumes of trade and assumed the adoption of the Panel's recommended conditions. Nonetheless, Professor Morris calculated, there was a moderate likelihood (30-70%) of a scenario 4 outbreak within five years, on a 50th percentile basis and a high (70-100%) likelihood on a 95th percentile basis. (b) Problems with the computer based model 194 Counsel next submitted that the 'computer based model substitutes its output for consideration of the statutory question'. As counsel noted, the ultimate question for the Director (under s 70 of the Proclamation) has two components: 'what is the level of quarantine risk for PMWS with the measures in place' and 'is that risk acceptably low?' Counsel said the computer based model prevented these questions ever being properly addressed and answered. Counsel noted the model had four critical steps: '(a) release assessment - restricted and unrestricted; (b) likelihood of entry and exposure assessment - restricted and unrestricted; (c) consequences assessment - common to both; (d) overall risk.' 195 Counsel claimed a number of problems were inherent in each step in the model. Counsel may not be accurate in describing all the problems as inherent. Whether or not they are, the alleged problems require consideration in determining whether, as the applicants contend, they led the Panel (and so the Director) to fail to take into account all material considerations. 196 Counsel claimed the following problems: (i) at p 391, the IRAR offers a 'low' assessment of the likelihood that imported pig meat derived from an individual carcass would be affected by PMWS, but it does not provide similar information in relation to restricted risk. Counsel submitted: 'Thus, the probability of the introduction of PMWS into Australia given the measures, being an element of the level of quarantine risk, is not reported'; (ii) as a result, the IRAR does not indicate the likely number of PMWS affected carcasses that will enter Australia each year, if permits are granted in accordance with the Panel's recommendations. Professor Morris calculated the number of PMWS affected whole carcasses that would be imported each year, on an unrestricted risk basis, as lying between 50,000 (50th percentile) and 125,000 (95th percentile). He said the number would be higher if parts of carcasses were imported, as is present practice. He commented: 'Thus at the expected volume of trade, very substantial numbers of carcases of infected animals will definitely be imported each year, and the protection against infection becoming established in Australia will depend on the risk management measures (deboning, removal of major peripheral lymph nodes, and cooking or curing) proving successful in reducing the likelihood of a pig becoming infected to the very low level required for Australia's ALOP. There is no scientific evidence that any one of these will influence the likelihood of pigs being exposed to the agent of PMWS, and the risk management strategy relies for its calculated effectiveness in reaching the ALOP, on a high degree of reduction in the amount of waste generated, rather than on specific measures for preventing or controlling entry or survival of the agent.' (iii) Even in relation to the unrestricted release assessment, the IRAR is problematic. Counsel said: 'Very large intervals are specified for some likelihoods - e.g. moderate spans 30% - 70%. Other intervals are very small. Instead of a best estimate being made on the R question, which is essentially a single question, the experts have to put their views into the very broad moderate box as they do for R1, R2 and R4, and then permit the computer to run its simulation and produce a supposed outcome of low (5% - 30%). In the process, the experts have been constrained in a way that what is not reported is their best estimate, plus an appropriate probability distribution on the single question of the likelihood that meat arriving in Australia has come from a carcass which contains an infectious PMWS agent.' (iv) The IRAR provides information only at the 50th percentile. This is despite the fact that the IRAR (at p 27) states that likelihoods assigned to the pathway steps considered in the analysis were 'modelled using the qualitative descriptors described in Biosecurity Australia's Guidelines for Import Risk Analysis' located at a particular website address. The reference is to the 2001 Guidelines. This document recommends, as a rule, reporting of the 95th percentile (see para 99 above). The IRAR gave no reason for the Panel's decision to depart from this rule. Counsel noted that Professor Gardner agreed with the desirability of reporting the 95th percentile but had said the IRAR could be defended if what had occurred was that very conservative estimates were taken at each of the input levels; that is, the estimates were more conservative than the true best estimates of the Panel. Counsel commented: 'The difficulty is that the Report does not state that it has modelled the experts' opinions at the input levels on the basis that they would be more conservative than their best estimate. Some are clearly not conservative e.g. R5 and R6. That being so, the reason from departing for [sic] the 95th percentile has not been given.' (v) uncertainty arises out of 'many subjective judgments being made based on incomplete, inadequate or uncertain data'. 197 In relation to the likelihood of entry and exposure (the L factor), counsel submitted: (i) the IRAR does not disclose any reasoning for the assumption that the output of the R calculation (relating to the importation of a single PMWS-affected carcass) equals L1 (the likelihood that a waste unit is PMWS-affected). Professor Gardner accepted that the L1 value might be higher, if carcasses were combined, as in a smallgoods factory; (ii) the IRAR does not disclose the assumed L1 after the restricted risk conditions are applied; (iii) the overall unrestricted L factor is merely described as 'high', notwithstanding that more precise data is available (99% in one case); (iv) the number of waste units was calculated by reference only to an informal survey of a limited number of people who worked in particular offices in Canberra. Counsel suggested this survey had no validity. 198 In dealing with consequences, counsel made the following points: (i) the four scenarios have been defined in such a way as to exclude the possibility of exposure and spread of PMWS beyond a single pig producing region. This is despite the overseas experience noted in the IRA. Counsel said the consequences of a more extreme scenario were not assessed in the IRA; (ii) the likelihoods are reported merely by reference to the probability distributions; for example, 'high' rather than the true assessed probabilities (such as 98% or 99%); (iii) the assessment table and decision rules have been so structured that, unless an outbreak has an impact at national level, it is virtually impossible for consequences ever to be classified as greater than 'low'. This is the case regardless of the impact at, say, State/Territorial level. The result of consequences being assessed as 'low' is that the overall risk cannot be assessed as higher than 'low'; (iv) because the Panel was confined to ticking boxes, no assessment was made of the real likely consequences of an outbreak of PMWS - 'pigs dead, jobs lost, income lost, farms closed'. Counsel noted that Professor Gardner said this type of data is 'usually central to a risk analysis' (see para 110 above). 199 In relation to assessment of the overall risk, counsel made five points: (i) because of the rules concerning national and State/Territory consequences, the overall risk can never be assessed as greater than low. In oral submissions, Mr Gleeson said: 'What happened here … is that, instead of answering the question which is what is the probable extent of harm, which requires you to say what's likely to happen, how many pigs are likely to die, how many farms will close, how many jobs will be lost, what income will be lost, are there any countervailing benefits perhaps, then putting some assessment on whether that is something which you are prepared to accept the risk of, … they've instead asked this question. This totally irrelevant question under the statute. Is it felt at national level or state level? Where in this statute is the decision-maker told to do that?' (ii) once the L factor falls below 30%, the table required the conclusion that the risk is 'very low'. However, Professor Gardner could not support a threshold as high as 30%; he said he 'might be happier' with a 5% threshold. Professor Gardner did not know any literature that supported a figure as high as 30%; (iii) the definition in s 5D of the Act does not authorise a determination of the level of quarantine risk by a simple multiplication of probabilities, nor does s 70 of the Proclamation contemplate a determination made in this way; (iv) the IRAR contains no consideration of benefits. Counsel for the applicants said this circumstance seems to amount to an acceptance by the Panel that there were no benefits to be weighed against the import risk. They went on: 'If this is to be taken at face value, provided there is some appreciable risk of entry and exposure, and provided the consequences are of some real significance (even if at a state level), then the absence of any countervailing benefits in the calculus mean that the risk could never be described as acceptably low under section 70. On the other hand, if benefits are being brought implicitly into the model, i.e. the benefits of free trade being built into the model so as to require under Table 10 the acceptance of any risk described by it as "very low", then the model is not transparent. It is not exposed to the decision maker the calculus of the costs and benefits which must be considered in order to form a view about risks which are acceptably low.' (v) because of differences in the incidence of PMWS, a generic report was insufficient; there needed to be an assessment in relation to each country from which pig meat was likely to be imported into Australia. (iii) Improper exercise of power - unreasonableness 200 Counsel for the applicants contended that, even if all their above arguments are rejected, the uncontradicted evidence of Professor Morris is that the result of the issue of permits in accordance with the IRA Decision will be a 'scenario 4' outbreak of PMWS in small commercial piggeries. There is a moderate chance of a scenario 4 outbreak in the first year, on a 95th percentile basis. Even on a 50th percentile basis, there is a moderate chance within five years; on a 95th percentile basis, a high chance within that period. 201 In their written submissions, counsel for the applicants said: 'The First Respondent does not, and cannot, refute those calculations as the outcome of his Department's own model. The First Respondent was not told of those calculations, because the model reduced the risk output to a single description "very low", combining all exposure groups and all scenarios. He was not told anything about the prospects of scenario 4 for each scenario group (after measures). He was not told anything at the 95th percentile. He was not told anything for a period beyond a single year. Perhaps worse, he was not told why he was not being told those matters. His ignorance was uninformed. On the balance of probabilities, the effect of the Decision is that, within a small number of years, there will be an outbreak of PMWS at the level, at least, of "scenario 4". If one were conservatively following the rule expressed in … the September 2001 Guidelines …, using the 95th percentile in a recognition that all models are at least to some extent imperfect representations of reality, then there is a chance of between 30% and 70% that an outbreak will occur within 12 months.' (Original emphasis) 202 Counsel noted that Professor Morris, Professor Gardner and Dr Allan had all agreed that a study could have been done to see whether feeding PMWS affected meat to healthy pigs led to infection. (iv) Reviewability of the two decisions 203 Finally, counsel for the applicants put submissions regarding the reviewability of the IRA Decision. They put two submissions: first, the IRA Decision is itself reviewable; second, the Permit Decision was made pursuant to, and by reason of, the IRA Decision, so its lawfulness cannot be assessed without reviewing the IRAR itself. 204 In relation to reviewability of the IRA Decision itself, counsel for the applicants accepted that the IRA Decision was a policy determination but they disputed the respondents' claim that it did not affect legal rights. They took as an example importation of pig meat from the United States. Prior to the IRA Decision, it was not possible to import United States pig meat into Australia; one effect of the IRA Decision was to enable the grant of permits for the importation of United States pig meat. 205 Counsel said the fact that the IRA Decision is a substantive decision susceptible to judicial review is 'further confirmed by the facts that: '(a) it is the outcome of a lengthy process, including extensive consultation; (b) the process also included a merits appeal to the Import Risk Analysis Appeals Panel (decisions so lacking in substance that they are not reviewable are rarely the subject of merits appeals); (c) the decision was notified to the World Trade Organization …; (d) the decision was the subject of a Quarantine Alert … and is notified on the AQIS database ICON … ICON is "AQIS's import conditions database" where importers can access information about Australian import conditions.' 206 In relation to their second argument, counsel said: '[I]t is plain that the decision to grant the particular permit to the Second Respondent was made pursuant to the [IRA] Decision. This is a clear case. The Second Respondent's application would necessarily have been rejected prior to the [IRA] Decision. After the [IRA] Decision, the only real questions for the delegate were: (a) whether the USA would be able to meet the conditions in the [IRA] Decision (including Animal Health and Public Health requirements); (b) whether there was any reason not to adopt conditions contained in the [IRA] Decision; and (c) the impact of the present proceedings.' 207 Counsel argued that, if the IRA Decision is unlawful (or, alternatively, represents a policy contrary to the Act), the Permit Decision must be set aside. They submitted that, '[o]n uncontroversial information now before the Court, it is clear that no decision maker rationally approaching the task required by the Act could have made' either the IRA Decision or the Permit Decision. The respondent's submissions (i) The nature of the IRAR 208 Counsel for the respondent urged that the IRAR 'be understood as the assessment of a group of scientific experts, using the model as a tool for organisation and comparison'. Counsel said it is clear that the PMWS technical working group and the Panel 'exercised their own judgment as to whether the output of the model represented a sensible result'. The evidence cited in support of the last proposition was an email of 19 February 2004 from Geoff Gard (apparently a member of the technical working group) to Dr Martin in which he agreed with a summary she had made of a teleconference on 11 February 2004 and commented: 'I feel that you have covered the possibility that the cause may be a virulent strain of PCV2 with your references to a different strain of virus, though Chris may prefer you to mention this specifically.' 209 Counsel argued: 'At each stage, the analysis had to deal with the probabilities of events. This was done in one of 2 ways. Where data allowed, probabilities were modelled as distributions with characteristics that matched the level of confidence in those probabilities. Otherwise the analysis used six standardised descriptors of probability, each of which was assigned a numerical range, and gave events that were almost certain a likelihood of 1. Where calculations needed to be done using these probability distributions or numerical ranges, a spreadsheet-based simulation model was used which performed the calculation many times ("iterations") using values randomly selected from within the relevant ranges. The result was a range of probabilities which could then be interpreted by the Panel.' (Footnote omitted) 210 Counsel did not identify any occasion in the IRAR upon which an 'almost certain' event was given a likelihood of one. As emerged from the oral evidence, there were at least two occasions (R5 and R6) when values of 98-99% were treated only as 'high'. (ii) Reviewability of the decisions 211 After summarising the material in the IRAR, the respondent's counsel dealt with the issue of reviewability. As already indicated, they agreed that the Permit Decision is reviewable. They accepted that the IRA Decision 'gives rise to an imminent likelihood' that the policy embedded in it 'will be taken into account in making other decisions to grant permits' under s 39(1) of the Proclamation. They conceded that, if it were shown that a decision that had been made under s 39(1) after taking that policy into account 'must necessarily be made in excess of jurisdiction then prohibition would at least in theory be available'. However, they said, the circumstances of the Permit Decision demonstrate it cannot be assumed that the policy 'has been or will be mechanistically applied by decision-makers'; it cannot be assumed that the policy in the IRAR 'will not be subjected to continuing revision as further information may come to light'. In that regard, counsel referred to the evidence (paras 164-169 above) concerning the actions of Dr Carroll and the Panel immediately before the making of the Permit Decision. 212 Counsel submitted that the IRA Decision is not susceptible of review under the ADJR Act; it is not a decision 'under an enactment'. They said: 'Unless and until it is taken into account in the making of a decision under s.39(1) of the [Proclamation], it amounts to nothing other than a formal statement of intent.' (iii) Was Agent X properly considered? 213 Counsel for the respondent saw the question whether Agent X was properly considered as something fundamental to several of the applicants' argued grounds of invalidity. Counsel claimed it had been properly considered and referred to a passage (at page 386 of the IRAR) in which the Panel noted it is generally considered that PCV2 is an essential, but not sufficient, factor inducing PMWS and listed possible co-factors (infectious and non-infectious). Counsel then referred to the debate, in the expert evidence, as to the most likely co-factor or co-factors. They concluded, accurately, that 'there is, as yet, no experimental data to support the existence of an Agent X'. 214 Counsel for the respondent identified two arguments of their opponents: 'First, it is said that it was irrational to proceed to make a decision as to the probability of harm without making any attempt to identify Agent X or to form a view as to whether, and if so to what extent, Agent X might be present in the muscle after cooking, deboning and removing the major peripheral lymph nodes. In particular, it is said that there was no basis for concluding that such measures would reduce R4 from "moderate" to "low" and L2 from "moderate" to "very low". Secondly, it is apparently said that it was irrational or unreasonable to proceed to make a decision at all without first commissioning a study at least to determine: (1) whether and if so to what extent PCV2 might be present in the muscle after cooking or curing; or (2) perhaps, to determine whether and if so to what extent pigs fed with PMWS affected meat might develop PMWS. Apparently in both respects, the way in which PMWS was dealt with is contrasted with the way PRRS was dealt with.' (Footnotes omitted) 215 Counsel for the respondent dealt with the first argument by saying: 'there is nothing to suggest that a conclusion that PCV2 remains present in muscle after cooking or curing would be likely to result in more conservative values for R4 or L2. It is common ground that the current state of scientific knowledge does not permit the conclusion that cooking or curing will reduce PCV2 in muscle if it is there. First it was assumed in the Final IRA that PCV2 was present in muscle and, secondly it is, and always was, clear that the Final IRA placed little weight on cooking and curing as a means to inactivate the PCV2 virus.' (Footnotes omitted) 216 Counsel submitted it is apparent from the IRAR that the Panel's reasoning took into account 'the possibility that there may be an Agent X and that it may be present in PMWS affected meat'. It was wrong to suggest that a judgment as to R4 or L2 (and ultimately as to the probability of PMWS being introduced through imported pig meat) could not prudently be made unless Agent X had first been ruled out or identified. It was for the Panel to determine what weight should be given to the possibility of Agent X. 217 Anyway, counsel argued, it was open to the Panel to perform its analysis on the assumption that a virulent strain of PCV2 is responsible for PMWS and then to ask whether different risk management measures would be needed if an unknown disease agent is involved. Counsel said: 'Proceeding on the assumption that a virulent strain of PCV2 was responsible for PMWS, the Panel first modelled each of the R and L factors. Subsequently, it explicitly found that the removal of the major peripheral lymph nodes and deboning could reduce R4 to "low" and L2 to "very low". It did so on the basis of an explicit assessment that "[i]n persistently infected pigs levels of virus in tissues are likely to be low". This must be read against the introductory observation that "[l]ymphoid tissues are the primary target tissues for PCV2, as for many other viruses", a statement borne out by the scientific literature which had already been extensively surveyed both as to the pathology and hardiness of PCV2.' (Footnotes omitted) 218 Counsel also noted the Panel's comment (at p 747 of the IRAR) about the hardiness of PCV2 and, therefore, that 'if an unknown disease agent was involved in PMWS, the risk management measures requiring removal of bone, major peripheral lymph nodes, head and neck and cooking and curing would act to reduce the risks associated with that agent'. 219 Counsel argued that the applicants were not entitled to make their second criticism regarding the lack of experimentation. They said the criticism had not been pleaded and was, therefore, not addressed in the respondent's evidence in chief. In any event, they said, the evidence goes no further than to affirm that such an experiment would be possible; its practicality and utility were not explained in the evidence. Counsel added: 'Again, there is nothing to suggest that a conclusion that pigs fed with PMWS affected meat might develop PMWS would be likely to result in more conservative values for R4 or L2. It ought be noted in this respect that for PRRS (where an experiment was conducted and did show the PRRS virus to be present in the muscle) the value of R4 chosen was "moderate" (the same as for PMWS) and with the removal of the lymph nodes was reduced to "low" (again, the same as for PMWS).' (Footnotes omitted) (iv) Was the methodology flawed? (a) annual risk 220 Counsel for the respondents argued the relevant issue 'is not so much annualised risk, but the volume of imports considered'. The overall risk 'depends on the figure predicted for the volume of trade'. They said: 'To assert an error of law, the Applicants must establish that s 70 of the Quarantine Proclamation required the Panel to consider the level of quarantine risk by reference to a volume of imported pig meat, greater than that adopted. The same test applies in relation to relevant considerations: administrative action can only be ultra vires on this ground "if a decision-maker fails to take into account a consideration which he is bound to take into account in making that decision: (Minister for Aboriginal Affairs v Peko-Wallsend Ltd ("Peko Wallsend") 162 CLR 24 at 39 (Mason J). Inevitably, the Applicants have failed to find any such legal obligation, because there is none.' (Counsel's emphasis) 221 Counsel pointed out that 'any likelihood must be expressed as a rate (per tonne, per year, per century etc)'. They argued 'there were good reasons to assess and describe likelihoods in annual terms'. They added: 'Likelihoods were thereby expressed for each year, not only for one year. To emphasise the higher percentages that emerge when probabilities are expressed over longer periods is, in substance, merely to engage in a policy debate as to what level of risk should be regarded as "acceptably low".' (Counsel's emphasis) 222 After referring to the debate as to the significance that ought to be accorded to the fact that PMWS outbreaks had not been recorded in Australia, despite importations from Canada and Denmark, counsel said: 'In carrying out the risk assessment, the Panel assumed the existence of an unknown and unidentified infectious agent: it did not assess the probabilities of such an agent existing. However, historical information as to the absence of any outbreak in the past, together with doubts expressed by the experts as to the existence of any such unknown agent, would, no doubt, need to be taken into account also if a longer period (or greater volume of imports) were to be addressed.' (b) use of a computer-based model 223 Counsel for the respondent accepted that the Panel's reasoning process incorporated a computer-based calculation, and this calculation was based on application of certain principles or rules. However, they said, 'the rules were not "substituted" for anything: they were the mechanism by which the assessment was carried out. If it is said that the assessment did not address some element contained in s 70 of the [Proclamation], that element must be identified by reference to the section'. 224 Counsel then dealt with the applicant's references to s 5D of the Act: 'The specific complaint [of counsel for the applicants] is that s 5D of the [Act] "does not authorise" a simple multiplication of probabilities as the determination of the level of quarantine risk. Section 70 of the [Proclamation] authorises (and requires) the Director to consider the level of quarantine risk, in circumstances to which it applies. Section 5D of the Quarantine Act defines a level of quarantine risk as a reference to certain probabilities. Taken together, the provisions authorise (and require) the consideration of certain probabilities. If the probability of an event can be assessed by multiplying the separate likelihoods in relation to independent components, no doubt that may be described as "a simple multiplication" (although the phrase hardly does justice to the process) and accordingly it is authorised by ss 70 and 5D. The level of "risk" may properly be assessed, in accordance with s 5D and the combination of the probability of an event and an estimate of its consequences. That was precisely the task undertaken by the IRA. The complaint … is that such a "simple multiplication" could not answer the question of what is "acceptably low" for the purposes of s.70 of the [Proclamation]. That may be accepted as a true statement: however, it demonstrates the weakness, rather than the strength, of the Applicants' case. What is "acceptably low" is a matter of evaluative judgment, to be made by a Director of Quarantine, on the basis of the relevant materials.' 225 Counsel for the respondent argued that, as a matter of principle, it is not necessary to assess further the specific complaints of the applicants regarding the IRAR. They commented, however, that 'lack of transparency' in relation to some matter does not constitute either error of law or a failure to take into account a relevant consideration. Counsel also defended the Panel's use of uniform probability distribution and broad ranges of likelihood. Counsel justified adoption of the 50th percentile by reference to Mr Vose's statement that 'there is no right or wrong to the selection of the percentile to use: it is a matter of choice'. (c) benefits 226 In relation to the IRAR's non-consideration of benefits, counsel said: 'The complaint in relation to this particular is twofold: first it is said that no proper assessment of risk can be made unless benefits of importation are taken in, as part of a balancing exercise. The legal justification for this conclusion is not apparent: there is no reference to such a balancing exercise, or the need to calculate the benefits of importation, in ss.39 or 5D of the Act, nor in s 70 of the [Proclamation]. Nor does any such justification appear from the statutory history or context, set out in Part 2 above. An assessment of quarantine risk requires an assessment of the probability of introduction, establishment or spread of the disease, the causing of harm and the probable extent of the harm. There is no reference to countervailing benefits. Nor is that concept inherent in the nature of "quarantine". Nor does the fact that an evaluative judgment of the acceptability of the level of risk is required necessarily import any balancing exercise.' (d) permit specific consideration 227 In response to the applicant's complaint that permit decisions should take account of the likely disease-inducing agents in particular countries, counsel for the respondents made two essential points: first, 'it cannot be wrong to consider the effect of decisions over a relevant period, cumulatively'; second, there is no evidence there was material before the Panel or Dr Carroll that suggested the identity of the particular country of origin was a material factor. Similarly, there was no evidence that it would matter whether there were two or three viruses, 'the hypothetical characteristics would be assessed for each, with the same result'. (v) Unreasonableness 228 Counsel for the respondent treated this ground as coming down to the question whether the Director had the information necessary for him to make an informed decision. In their written submissions, they summarised the applicant's arguments in this way: 'First, it is said that the way in which the IRA was written denied the Director information necessary to make a rational decision. The information said to be denied comprises the calculations performed by Professor Morris in par 10.8.1 of his report in reply which show the likelihood of a scenario 4 outbreak for each exposure group at the 50th, 75th and 95th percentiles over a one, five and ten year period. The calculations demonstrate that what was presented in the Final IRA as an overall "very low" annual risk really can also be presented as a moderate probability of a scenario 4 outbreak occurring at the 50th percentile over a 5 to 10 year period and a high probability of a scenario 4 outbreak occurring at the 75th and 95th percentiles over the same 5 to 10 year period. It is complained that the Director was told nothing about the probability of a scenario 4 outbreak occurring or of the 95th percentile and was not informed that "on the balance of probabilities" the effect of the Determination "there will be an outbreak of PMWS at the level, at least, of scenario 4.' 229 Counsel for the respondent offered what they called 'two complete answers': 'The first is that, at the level of principle, the Director is entitled to rely on his Department "to undertake an analysis, evaluation and précis of material to which the [Director] is bound to have regard or to which the [Director] may wish to have regard in making decisions: (Peko-Wallsend at 65). Here, the Director was informed in the body of the Final IRA Report of the material taken into account, of the methodology employed by the Department in evaluating that material and of the results of that evaluation. The methodology about which the Director was informed did not provide for the separate reporting of the probability of a scenario 4 outbreak occurring and did not provide for reporting of the 95th percentile. Whether or not it may have been preferable for the Director to have been told that additional information in the Final IRA, it simply cannot be said that it was irrational for the Director to proceed to make a decision without that additional information. The evidence in chief of Professor Morris does not go so far as to say it would have been irrational to proceed to make a decision without that additional information and both Professor Gardner and Mr Vose denied in cross-examination that it was irrational.' (Footnotes omitted) 230 Counsel's second 'complete answer' was to point to material in the Annex to the IRAR. They said information similar to that provided by Professor Morris: 'was presented in the form of tables prepared for APL, not by Professor Morris, but by the CSIRO. Those tables formed an annexure to a submission by APL on the Draft IRA. By reference to that information, the APL submission specifically made the point that the likelihood of one or more outbreaks of PMWS occurring over a ten year period was 99%. The same point was then repeated by APL in its appeal to the IRAAP and was specifically referred to by the IRAAP in its report to the Director.' (Footnotes omitted) 231 Counsel dealt with the 30% demarcation agreement as follows: 'Next it is said that even on the information contained in the Final IRA it can be seen that the likelihood of entry and exposure for small commercial piggeries is between 5 and 30% and that "there is no rational basis on which a chance of up to 30% of such significant consequences could be described as acceptably low". But that is precisely the result rationally arrived at by application of the methodology. If the methodology is sound, the result cannot be irrational. If the methodology is left entirely to one side, there is no basis whatsoever for saying that it is perverse to label a risk based on a 5 to 30% chance of entry and exposure for small commercial piggeries as "very low". The matter is one of subjective evaluation of the combination of probability and harm.' (vi) Relief 232 Finally, counsel for the respondent made some comments about the form of any relief that might be granted by the Court; a matter on which, if necessary, they sought to be heard. Applicants' submissions in reply 233 The applicants made a number of points in their reply. I will mention only those that are truly new. 234 First, counsel disputed that their complaint about the Panel's failure to conduct any experiment concerning the effect of feeding PMWS affected meat to healthy pigs was an unpleaded afterthought. They argued it was covered by para 53 of their Points of Claim. Perhaps more pertinently, they pointed to a passage at paras 10.3.5 to 10.3.8, in Professor Morris' first report, served on the same day as the Points of Claim, as follows: 'The more serious issue is whether removal of lymph nodes is likely to substantially modify the concentration of PMWS infectious material in the pig meat. As in earlier discussion, the IRA, instead of considering PMWS, reverts to considering PCV2, which is a flawed assessment. Moreover, as the IRA itself states, it is unknown whether the virus causing PMWS is present in meat, and this is a critical piece of information in assessing the adequacy of the proposed measures. In the case of PRRS virus, Australia funded a study which showed (contrary to earlier untested assumptions by risk analysts) that PRRS virus survived in meat and would potentially be infectious in product imported into Australia and New Zealand. Both countries then tightened their risk management procedures in the light of this information. Although the virus of PMWS cannot yet be isolated, it would be possible to conduct an equivalent study using meat from animals in the early stages of PMWS as the source of potentially infectious product. Given Australian experience with the PRRS case, it is most surprising that this option is not even considered in the IRA. In my view, it would be a very valuable study to conduct, and would be of great interest to a number of countries.' 235 Counsel for the applicants said that, far from being unable to meet such a case, Dr Allan responded expressly to those paragraphs by agreeing it would be possible to conduct experiments to determine 'whether the virus causing PMWS is present in meat'. Dr Allan expressed surprise 'that Professor Morris and his colleagues in Massey University have not already conducted these experiments in respect of their proposed Agent X'. 236 In their reply counsel commented: 'The experiments are straightforward, not worthy of a Nobel prize, and although they may not determine the identity of Agent X, they would enable an informed assessment of the questions raised by s70: the probability of introducing PMWS into Australia, and the impact on that probability of imposing conditions.' 237 Counsel responded to the respondent's submission that there was nothing to suggest that an experiment would have resulted in more conservative values for R4 or L2. They said: 'That simply misses the entire point of the argument. Without conducting the relevant experiments, there was no rational basis for the First Respondent to form a view about what R4 and L2 would be, especially after the restricted measures were adopted. If the Panel's reasoning were to adopt a virulent strain of PCV2 as the surrogate for the possible unknown agent … then the First Respondent, absent experimentation, simply did not have any data upon which to conclude: (a) what would be the level of the virus in the muscle? (b) whether that level of virus would be likely to survive in the muscle to the point where pig meat waste was available for consumption by a healthy pig in Australia? (c) accordingly, what was the real chance of PMWS being established in Australia through this mechanism? Nothing in the First Respondent's submissions … addresses the critical question: whatever reduction in the amount of the virus (whether it be a virulent strain of PCV2A or agent X or something else) may occur by removal of the lymphoid tissue, it says nothing about whether the remaining level of virus in the muscle has a sufficient likelihood of causing PMWS to establish and spread in Australia.' [Original emphasis] 238 Counsel for the applicants repeated their denial that the IRAR was conservative. They emphasised Mr Vose's concession that the effect of treating R5 and R6 as 'high' rather than as '1' on the basis they were near certainties, was to understate the R probability by about 30%. Counsel added: 'The same is true of R4. Although R4 was submitted [by the respondent] to be a "conservative element", … a better description is that it is a "conservative guess", because there is no factual foundation in the science at all. It is to be recalled that the question R4 is answering is: assuming a pig is infected with the infectious agent, what is the chance that the meat from that pig that is harvested for export will contain the infectious agent? There is no reason to think that there is only a "moderate" 30% to 70% chance that the agent would be present. A conservative estimate would not have reduced the overall likelihood to any extent in respect of this factor; but the IRA Report halves the total risk that is computed by the Monte Carlo simulation by reason of assigning a "moderate" label to this input (and then using the 50th percentile of the output distribution).' (Original emphasis) 239 In relation to unreasonableness, counsel said: 'The First Respondent does not cavil with the applicant's proposition that on the information obtained in the IRA, by extrapolation, there is a likelihood of entry and exposure (even at the 50th percentile and limited to one year's volume of trade) for small commercial piggeries of between 5% and 30%, carrying with it the very significant consequences of Scenario 4. That is a risk which a rational decision maker could never consider to be "acceptably low". What is the First Respondent's answer? Only that "that is precisely the result rationally arrived at by application of the methodology". There could be no clearer indication that the methodology was never apt to answer the statutory question.' (Original emphasis) Conclusions (i) IRAR problems (a) introduction 240 In their closing submissions, counsel for the respondent referred to a comment made by Mr Vose in his initial report: 'In my view, Biosecurity Australia has done a remarkable job in achieving the aims put forward by the SPS agreement and the OIE Code. I use the term "remarkable" because, as far as I am aware, no other country has produced an import risk analysis methodology that has explicitly stated how it will evaluate the consequence of a disease introduction nor matched their evaluation to a risk management action that achieves the ALOP. Most countries, if they publish their import risk assessments at all, consider only the probability of disease introduction.' 241 As mentioned at para 158 above, in oral evidence Mr Vose said 'no other risk analysis I've seen has had the guts, if you like, to address the consequence component of the risk assessment'. 242 From these and other snippets of evidence, and from the opening sections of the document itself, I infer the IRAR was intended by its authors to achieve new standards of comprehensiveness and excellence in import risk analysis; and perhaps to provide a model that might be used in relation to other commodities and even by other countries. 243 Against that background, it gives me no pleasure to say I have reached the conclusion that there are serious problems about the IRAR, at least in relation to PMWS. The effect of those problems, in relation to legal invalidity, is another matter. 244 The IRAR was written to enable the Director to adopt a general policy that delegates would take into account, and usually apply, in determining particular applications under cl 39 of the Proclamation to import pig meat from other countries, including countries where PMWS is commonplace. In determining whether to grant that permit, the delegate is required, by cl 70 of the Proclamation, to consider, first, 'the level of quarantine risk if the permit were granted' and, second, what conditions are necessary 'to limit the level of quarantine risk to one that is acceptably low'. 245 It is clear that PMWS has had devastating effects on pig herds in many countries. PMWS currently costs the European Union about 600 million Euros per year. It is common ground that, so far, Australia is free of the disease. Yet the undisputed evidence of Professor Morris is that, if permits are granted in accordance with the policy embodied in the IRAR (and with full application of the Panel's recommended conditions), there is a 'high' risk - that is, an over 70% chance - at the 75th and 95th percentiles ('moderate' at the 50th percentile) that, within five years, PMWS will have spread to the general population of Australian domestic pigs, including medium-large piggeries. The IRAR itself, looking further ahead, showed a 99% chance at the 50th percentile (100% at the 95th percentile) of one or more outbreaks over ten years (Table 1, Annex C in the Annexes volume). In other words, under the policy that has now been adopted by the Director, an outbreak of PMWS within ten years is a virtual certainty. Indeed, according to Table 1, the problem may be more imminent. The Table shows a 38% chance at the 50th percentile (98% at the 95th percentile) of an outbreak as early as the first year. 246 An assessment that such a risk is 'acceptably low' seems to me bizarre, especially having regard to concern expressed by successive Australian governments about maintenance of high quarantine standards. Intuitively, one feels, there must be something wrong with the Panel's assessment of risk. 247 Any discussion of defects ought to take place against the background of the stated purpose of the IRAR. At page 12 of the IRAR, the Panel stated the IRAR's purpose was 'to deliver a policy recommendation to the [Director] that is characterised by sound science and by transparency, fairness and consistency'. 248 I have already detailed the experts' evidence on the question whether this purpose has been achieved in relation to PMWS. I have also summarised counsel's submissions on the matter. This makes it possible for me to express my conclusions relatively briefly. 249 I believe the bizarre result flows from three aspects of the methodology adopted by the Panel and from an unsubstantiated final step, taken by the Panel in converting the 'low' unrestricted risk to a 'very low' restricted risk. 250 The three aspects of methodology were the Panel's decision to limit its assessment to annual risk, rather than long-term risk; the use of broad probability distribution bands (except at the lower end of the spectrum); and adoption of the 50th percentile rather than the 95th percentile. Each of these decisions has a profound effect on the outcome, yet none of them was explained in the IRAR; so much for transparency. None of them was convincingly defended by a witness. (b) annual risk 251 Mr Vose defended the Panel's decision to consider risk only on an annual basis by remarking that assessments of financial feasibility are usually made on a 'first year' basis. Expenditure that will be incurred, or income that will be received, in the second and later years is usually not brought into a feasibility assessment at its actual estimated dollar value but by reference to the present (that is, year one) worth of the estimated number of dollars. In the case of a financial calculation, this approach is readily justifiable; monies required for investment in later years can earn interest in the meantime; the delay that will occur before revenue from later years is received may make it necessary for the proponent to borrow money in the meantime. 252 However, it is difficult to see the sense of such a limitation when one is considering what long-term policy ought to be adopted to guard against introduction of a new disease into Australia. In such a case, what is important is the effect of that policy over its whole life, not just its effect in the first year. The effect of the policy is proportional to the total volume of product imported into Australia during the life of the policy. The policy recommendations of the IRAR were intended to be of indefinite duration.. 253 The decision to introduce rabbits was possibly the most disastrous importation decision ever made in relation to Australia. However, if somebody had made a risk assessment before that decision, with a one-year perspective, importation would easily have been justified. The assessment would have found that, within the first year, rabbit importation would result in increased availability of fresh meat without significant agricultural damage. 254 Factual conditions vary over time. For that reason, it would have been artificial for the Panel to have taken a very long period - such as the period of a century mentioned by Mr Vose: para 143 above. However, perspective might have been gained if, as Professor Pettitt suggested, the Panel had considered the situation over (say) ten years, assuming a continuation of current conditions with such variations as were indicated by present knowledge. (c) the probability bands 255 The Panel's decision to use Monte Carlo risk simulation was not itself problematic. However, the evidence indicates that, if this technique is to be used, it is important to select bands of probability distribution that enable the simulation accurately to reflect the true estimates of the experts. A curious feature of the probability bands adopted by the Panel is that they include no less than three bands (very low, extremely low and negligible) for that portion of the probability spectrum that lies below 5%, but none for the top 5% of the spectrum; that is probabilities above 95%. As a result of this, the IRAR treated only as 'high' events that had been assessed as being 98% or 99% likely to occur. The effect of that step, in the Monte Carlo risk simulation, was that these events were treated as only 85% likely to occur. As Professor Pettitt pointed out, four events each given a likelihood of 85% result, in the Monte Carlo simulation, in a cumulative reading of 52%. In other words, a judgment by an expert panel that each of four events had a high (even 99%) chance of occurring would yield, via the computer, a conclusion that the happening of all four events was only slightly more probable than not. This result flows from adoption of a single probability band to cover all likelihoods in the range 70% to 100%. Unless the Panel were prepared to say a particular event was an absolute certainty (and I do not think they ever did this in relation to PMWS), the Panel had no option but to place that event in a probability band whose effective value might significantly understate the Panel's actual estimate of its probability. 256 Another concern about the IRAR's probability distribution bands is the width of 'moderate' (30% - 70%). There is a considerable difference between those extremes. A person who says there is a 30% chance of an event conveys a belief that the event is unlikely to happen, but concedes it may; there is a small chance of that. A person who speaks of a 70% chance intends the hearer to understand the person thinks the event will happen, though it may not. The Panel were obliged to disregard these nuances. A judgment by the Panel that lay anywhere between 30% and 70%, however close to one of the extremes, had to be called 'moderate' and was effectively treated, by the Monte Carlo simulation analysis, as a 50% chance. 257 One result of the Panel's decision to adopt a set of predetermined broad probability bands rather than actual numbers, or bands selected in relation to the particular event, was exposed in Mr Gleeson's cross-examination of Professor Gardner: see para 113 above. Demarcation points become critical. As Professor Gardner acknowledged, the effect of the rules adopted in the IRAR is that if the likelihood of entry and exposure is assessed 'at 31 per cent you grant the permit and at 29 per cent you don't'. 258 It surely is very difficult for any person, however expert, to distinguish between a probability of 29% and a probability of 31%. Yet under the methodology adopted in the IRA, that distinction is important. If the Panel opted for the lower of these two figures, in relation to any event, that item went into the analysis, effectively, at 17˝% (the mid point on the 'low' scale between 5% and 30%); if the Panel adopted 31%, it came in nearly three times higher at 50% (the mid point on the 'moderate' scale). Rules that enabled the Panel to feed into the Monte Carlo simulation their actual best estimate, be it 29% or 31%, would have ensured an outcome that better reflected its true opinion. Under such rules, if the Panel were unhappy about selecting a single figure, they could select the range of figures which most accurately reflected their view about the probability of that event occurring. The mid-point of that range would then be the effective input to the Monte Carlo simulation. That course is surely preferable to the Panel having to choose between two predetermined ranges whose common boundary might lie near the middle of their own assessed range for that event, and whose outer limits might be well outside their assessment. 259 Mr Basten rightly emphasised the expert qualifications of the members of the Panel and the technical working group. However, that emphasis points up the vice of the adopted methodology. The selected bands reduce the influence of the Panel's expertise on the IRAR outcome. Poorly selected probability bands may significantly distort the experts' true views. 260 None of the expert witnesses sought to justify the Panel's choice of probability distribution bands. The IRAR did not advance any justification. None of the Panel members gave evidence explaining their band selection. Professor Gardner was critical of the course taken in the IRAR. He told Mr Gleeson that, 'if there were hard data available and it [the L factor for feral pigs] could be modelled as anything but a uniform distribution between 70 and 100 percent then I would certainly model it with a different shaped distribution if that were more appropriate': see para 104 above. He also said he would take into account any available information about the actual percentage of an integer in the assessment: see para 105. 261 Professor Gardner would have tried to obtain numerical data wherever possible. He agreed with Mr Gleeson's suggestion (para 110 above) that, in assessing the consequences of a secondary spread of PMWS to a more general population of pigs 'in an ideal world', he would express a best estimate and a range in numerical terms. See also Professor Gardner's expression of concern (para 114) about 'the way it [the model] was done' and his wish for 'better numeric values to place on some of these inputs'. 262 Even Mr Vose, the expert witness who was most supportive of the IRAR, recognised the problem caused by its adoption of broad probability distribution bands: see paras 155 and 156 above. He pointed out that 'density falls sharply to zero at the minimum and maximum in an unnatural way'; there will always be 'a value that is on the cusp'. Indeed, he said, 'perhaps the only really good use of a uniform distribution is when you want to exaggerate your uncertainties and it allows you to do a sensitivity analysis'. 263 Of course, that was not the purpose of the uniform distribution in the present case. Its effect was to provide a result that the Panel may not otherwise have reached and to suggest a degree of certainty that the Panel may not in fact have possessed. (d) use of the 50th percentile 264 As counsel for the applicants pointed out (para 196 above), the IRAR claimed its analysis had been modelled in accordance with the 2001 Guidelines. In one important respect, at least, it had not. Those Guidelines recommend reporting the 95th percentile of output distributions. Although some 95th percentile data was included in the Annexes, the results stated in the IRAR itself, which results were those taken into the Monte Carlo simulation, reflected 50th percentile values. The IRAR gave no reason for this departure from the Guidelines. 265 Professor Gardner said that, if he had been doing the analysis, he would have 'done 50th, 75th and probably 95th percentile'; it is common practice to do this: see para 100 above. See also the evidence of Professor Pettitt mentioned at paras 86 to 88 above. 266 Perhaps the Panel's departure from the Guidelines and common practice might be justified if, as Mr Vose speculated, the Panel had knowingly adopted a conservative approach in all other respects. Although at the price of sacrificing transparency, such an approach might yield a result in which the Panel was able to have an acceptable degree of confidence. However, there is no indication in the IRAR, and no evidence to suggest, that the Panel consistently adopted a conservative approach. It certainly did not do so in relation to impact values for R5 and R6, where two assessments of over 99% effectively each became 85%. (e) the consequences table 267 Another valid criticism of the IRAR concerns Table 8, dealing with consequences. This table distinguishes between consequences on the national, State/Territory, district/region and local level. The distinction depends on geographical boundaries, not the scale of the envisaged disaster. Subject to two possible qualifications, under a table such as this, a plant disease that had the potential to wipe out all sugar cane growing north of a particular latitude in southern Queensland would be classified only as having 'State/Territory' consequences. On the other hand, a disease that wiped out the Australian cotton crop would have a 'national' effect; cotton is grown in both New South Wales and Queensland. 268 The first possible qualification to that comment is a statement at the top of p 64 of the IRAR: 'At each of the lower levels, an impact more serious than "minor" was deemed to be discernible at the level above'. Counsel differed as to the meaning of this statement. Counsel for the respondent argued that the sugar cane example would be treated as more serious than 'minor' and, therefore, as discernible at the national level, even though direct damage was confined to Queensland. Counsel for the applicants contended this interpretation is incorrect; the relevant sentence is governed by the previous sentence that '[t]he impact of a disease at a given level in more than one State/Territory, district or local area was considered to represent the same magnitude of impact at the level above'. Even if the respondent's construction is accepted, the damage would still be classified as 'minor', although with an E impact score rather than a D. 269 The other qualification concerns indirect consequences. Counsel for the respondent argued that a potential disease could be regarded as discernible at the national level if it would significantly affect the value of Australian gross domestic product and/or government revenues and expenditures. Perhaps this is so. However, such an approach involves a difficult question of degree. Whenever production is lost because of disease, there is an adverse effect on gross domestic product and on government revenues/expenditure. At what point does the indirect effect transform a State/Territory effect into a national effect? Why did the degree of PMWS impact implicitly assumed by the Panel fail to meet that threshold? The IRAR does not say. Indeed, there is nothing to suggest the Panel made any estimate of the losses, direct or indirect, likely to be sustained in a PMWS outbreak in Australia. Professor Gardner would have wished to have such data. He said it was 'usually central to a risk analysis'. 270 If the Panel are not to be criticised for failing to obtain detailed information about overseas PMWS losses, and then using this information to estimate the likely range of Australian losses, it is because their adopted methodology did not call for hard data. Rather, the methodology encouraged the use of nebulous descriptions like 'highly significant', 'significant' and 'minor', these being assessed by reference to geographical boundaries rather than the industry itself. (f) conversion of unrestricted risk to restricted risk 271 The most serious problem about the IRAR, in my opinion, arises out of its conversion of the assessed 'low' risk, before application of the recommended conditions, to 'very low' risk, after the conditions. This step was critical to the result of the study and the IRA Decision itself. Australia's ALOP is 'very low risk'. 272 I have previously quoted the crucial step in the Panel's reasoning. It is worth repeating: 'As PCV2 has been shown to result in persistent infections, has been isolated from many tissues, is a hardy virus and is likely to be transmitted orally, the Panel considered that if an unknown disease agent was involved in PMWS, the risk management measures requiring removal of bone, major peripheral lymph nodes, head and neck and cooking or curing would act to reduce the risks associated with that agent.' 273 The reasoning in that sentence may be summarised in this way: (i) PCV2 has certain characteristics, including being present in many tissues and being a hardy virus; (ii) Therefore, assuming an unknown infectious co-factor of PMWS, particular risk management measures 'would act to reduce the risks associated with that agent'. 274 There appears to be no connection between the first proposition and the second. It may readily be accepted, as Dr Allan stated at para 119 above, that the recommended risk management measures would have a tendency to reduce the risk posed by the assumed PMWS co-factor. This acceptance is a matter of common sense; it is not necessary to justify it by reference to PCV2. Any reduction in the volume of imported infected material must have a tendency to reduce the risk of infecting Australian pigs. The problem is that nobody can say whether the reduction would be great or small or sufficient to make a qualitative change in the risk. The Panel did not explain the basis upon, or the reasoning by which, it assessed the risk as having been reduced from 'low' (5% to 30%) to 'very low', that is, 5% or less. Dr Allan agreed with Mr Gleeson that, even after application of the recommended risk management measures, a carcass might still retain enough of the co-factor to cause PMWS infection: see paras 120 - 122 above. 275 Like everyone else, the Panel were unaware of the identity of the co-factor or co-factors that, with PCV2, cause PMWS. This was a huge problem, as the respondents' experts agreed: see, for example, Professor Gardner at para 115 above and Professor Ellis at para 133. 276 The Panel reacted to the problem by making the conservative assumption that at least one co-factor was an infectious agent, possibly a new virus or a particularly virulent strain of PCV2. That was a reasonable course to take. However, it left the Panel in the position of dealing with an assumed infectious agent of whose characteristics they were unaware. 277 PCV2 is known to be a particularly hardy virus. So the Panel made a judgment that any co-factor virus was unlikely to be more hardy than PCV2. Counsel for the applicants criticised the Panel for taking that course. Professor Ellis did not think such a judgment could safely be made: see para 132 above. I understand counsel's criticism; the Panel's judgment may be wrong. However, I think it was open to an expert body to make this judgment. 278 On that basis, it seems to me, the Panel was entitled to treat PCV2 as a surrogate for the presumed infectious agent; in the sense that risk management measures that will eliminate the risk of PCV2 infection being imported may reasonably be regarded as ensuring the assumed PMWS infectious agent also will not survive importation. 279 However, the Panel did not know what measures are necessary to ensure PCV2 infection will not survive importation. Deboning and removal of the head, neck and major lymph glands must help. However, Dr Allan said he did not know the extent to which these steps would reduce or eliminate the volume of PCV2 in the carcass; he had never researched the topic: see para 122 above. Nor was he aware of research by anyone else: see para 123 above. Professor Ellis said, if these steps were taken, he thought the odds would be long on survival of enough infectious agent to cause transmission. But he agreed not much virus was needed for transmission to occur; the virus need not be multiplying in the meat, it is enough 'if it's there and it remains alive in a small dose': see para 131 above. 280 Accepting the Panel's entitlement to use PCV2 as a surrogate, in considering what measures were necessary to reduce the risk of infection by the assumed co-factor to 'very low', the Panel had no basis for moving beyond what they knew about PCV2. If they did not know whether the recommended measures would be effective to reduce the risk of PCV2 infection to 'very low', reference to PCV2 provided no basis for selection of the restrictions necessary to guard against the assumed PMWS infectious agent. Of course, the Panel might have commissioned some research about the effectiveness of the proposed restrictions against PCV2, and used the result of that research in coming to a surrogate conclusion about PMWS. However, the Panel did not take this course. 281 Another step the Panel might have taken was to commission research, probably overseas, similar to that undertaken in relation to PRRS. Such research might have demonstrated that any infectious co-factor of PMWS (whatever it was) does not survive deboning and removal of the head, neck and major lymph glands. It may have demonstrated that the infectious co-factor would not survive cooking the meat at a particular temperature for a particular period of time. Even though such research may not have revealed the identity of the PMWS co-factor, its results may have enabled the Panel to devise conditions that would be effective against the infectious co-factor, whatever it was. 282 Mr Basten rightly said that the Panel was not under a legal obligation to commission research. However, the difficulty is that, in the absence of such research, the Panel lacked the information crucial to the final step in their reasoning; with the result that the step lacked rational foundation. 283 A more particular criticism of the Panel's reasoning, and conditions, is that nowhere did they explain what they meant by 'cooking', an option that was offered as an alternative to curing. The Panel noted that porcine circovirus 'has been reported to be stable at 70şC for 15 minutes'. Presumably, this statement was believed to be true of PCV2. Therefore, cooking to 70şC for 15 minutes would not affect the survival of PCV2. Treating PCV2 as a surrogate for the assumed infectious co-factor of PMWS, nor would it affect the survival of that co-factor. If cooking to 70şC for 15 minutes was intended to be enough to satisfy the condition, this constitutes a further logical flaw in the Panel's reasoning. If this was not intended to be enough, the Panel did not say what more was required. (ii) Consequences of these problems (a) the applicants' grounds of invalidity 284 Counsel for the applicants argued that the problems discussed above mean the IRA Decision is legally invalid. As it is clear, they said, that Dr Carroll took the IRA Decision into account in making the Permit Decision, the invalidity of the IRA Decision also invalidates the Permit Decision. 285 As I have mentioned, counsel invoked four grounds of legal invalidity. I will refer to them by reference to ss 5 and 6 of the ADJR Act. Section 5 lists grounds referable to a decision to which the ADJR Act applies. Section 6 deals with 'conduct for the purpose of making a decision to which this Act applies'. The applicants accept that the IRA Decision was not a decision 'under an enactment', so as to be caught by s 5 of the ADJR Act; but they argue that s 6 applies. They say the recommendations made in the IRAR constituted conduct for the purpose of enabling the Director and the Director's delegates to make decisions about the grant of individual permits, each of which decisions was a decision 'under an enactment'; that is, cl 39 of the Proclamation. (b) error of law and failure to take into account a relevant consideration 286 The grounds are error of law (s 5(1)(f) and s 6 (1)(f)), failure to take into account a relevant consideration (s 5(2)(b) and s 6(2)(b)), unreasonableness (s 5(2)(g) and s 6(2)(g)) and no evidence (s 5(1)(h) and s 6(1)(h)). 287 The applicants' Points of Claim gave the same two alternative sets of particulars in respect of each of the first two grounds. The first set of particulars (para 1 of the Points of Claim) reads: '(a) The respondent was required, pursuant to s70 of the Quarantine Proclamation, to consider the level of quarantine risk if a permit to import pig meat were to be granted by reference to the probability of any disease being introduced, established or spread in Australia and causing harm to animals, and to the probable extent of the harm. (b) The respondent failed properly to consider the level of quarantine risk that would be incurred in respect of PMWS in the event that a permit to import pig meat were to be granted. (c) PMWS is a disease which has been described in many countries, but not in Australia. (d) The virus PCV2 is a necessary but not sufficient factor for the presence of PMWS. (e) The additional factor necessary for the expression of PMWS is unidentified; it may be an unknown disease agent. (f) A consideration by the respondent of conditions designed to minimise the quarantine risk of PCV2 (a virus already present in Australia) does not discharge the obligation in s70 of the Quarantine Proclamation in respect of PMWS (or the unknown disease agent which may be involved in its expression).'