LL1-09 — Ch9 Antimicrobials as growth promoters: resistance to common sense#

Report: Late lessons from early warnings: the precautionary principle 1896–2000 (EEA Environmental Issue Report No 22, 2001), Chapter 9. Report pages: 93–100 (PDF pages 93–100; printed and PDF numbers coincide). Running text pp. 93–98; Tables 9.1 and 9.2 on p. 99; references pp. 99–100.

Reading record. I read the full text extract in order, through the final marker (PDF 100 / p. 100), and checked all eight PDF pages visually. The extract is faithful. The only problem is on p. 99, where the extract interleaves Tables 9.1 and 9.2 with the references; both tables were read from the page image. There are no boxes, figures or panels (the 2001 report has no panel commentaries). For standpoint and cross-reference only, I also consulted, in the same PDF, the author biographies (pp. 195, 198), the front-matter acknowledgements (PDF p. 6), and the editors’ synthesis passages that cite this case (pp. 170–182, 192). These are flagged as outside the section wherever they are used.

Authors and standpoint#

Authors. Lars-Erik Edqvist and Knud Børge Pedersen (p. 93). The chapter gives no affiliations. The report’s author annex does, and they matter:

• Edqvist (Professor, DVM, PhD) was Director General of the Swedish National Veterinary Institute (SVA), whose antibiotics department “has advocated prudent use of antimicrobials in veterinary medicine”. In 1995 the Swedish government appointed him chairman of the Commission on Antimicrobial Feed Additives, which evaluated the science “in relation to the temporary derogation from Community legislation that Sweden had when it became a member of the EU” and completed its task in 1997 (p. 195). That report is SOU 1997, the chapter’s most-quoted source (pp. 95–96). Its wording is reused in the chapter’s conclusions (p. 98).
• Pedersen (DVM, PhD, DVSc) had directed the Danish Veterinary Laboratory (DVL) since 1985. DVL set up research groups on resistance “particularly concerning antimicrobial growth promoters”, and Pedersen “initiated the formulation of a veterinary antibiotic policy and guidelines for prudent use of antimicrobials in veterinary medicine” (p. 198). From October 2000 he also directed the Danish Veterinary Institute for Virus Research (p. 198). DVL 1995 is cited as the scientific basis of the Danish avoparcin ban (p. 96). F. M. Aarestrup, lead author of three cited primary studies, was (I believe; not stated in the report) at DVL.

Standpoint. This is an insider account by senior government veterinary scientists from the two countries presented as frontrunners. Their institutions produced key evidence, and in Edqvist’s case the key official assessment, on which the story of vindication rests. The link is disclosed in the author annex but not in the chapter. The annex carries a general disclaimer that chapter views “do not necessarily reflect those of the institutions where they work” (p. 195, n. 14). The chapter itself says that, after Sweden obtained a temporary derogation in its EU accession negotiations, the commission was appointed “in support of the Swedish view” (p. 95). So the chapter gives direct expert knowledge, and it is also partly a defence of the authors’ own institutions’ positions.

Evident stance. Clearly pro-precaution. The title puns on bacterial resistance and institutional resistance to common sense. The arc runs warning → dilution → vindication, and Swann and the farmers are “vindicated by history” (p. 98). The tone is measured. The authors acknowledge production benefits (p. 93; Table 9.1). They concede that human medical use “undoubtedly is of more importance” for resistance in humans (pp. 97–98). They hedge some causal claims (“probably”, p. 94).

Panels. None. The report’s front matter lists peer reviewers for the whole report, including Wolfgang Witte, Stuart Levy and Richard Young (PDF p. 6), but does not say who reviewed which chapter. Witte co-authored Klare et al. 1995, one of the three studies cited as first evidence of vancomycin-resistant enterococci (VRE) in animals (pp. 96, 100). Outside the section: the author annex records that Jim Bridges, co-author of a different chapter, sat on SCAN in 1991–97 and on the EU Scientific Steering Committee from 1997 (p. 195). So the report as a whole included someone from inside those committees, but Ch. 9 does not draw on that perspective.

Absent or thin voices. The pharmaceutical industry appears only as an opponent or litigant (pp. 94, 96). The scientific case that growth-promoter use was safe appears only through one quoted scientist, Walton (1988, p. 94). (Their benefits are acknowledged by the authors, p. 93 and Table 9.1, and by the 1977 Swedish working group, p. 95.) The chapter gives no affiliation for Walton and does not present him as speaking for industry. The reasoning of the EU Scientific Committee on Animal Nutrition (SCAN) is summarised in two sentences (p. 96). Apart from the Swann Report itself, there are no UK government sources, and none on the dilution decisions. There are no US regulators and no economic data.

Section-by-section notes#

9.1 Introduction (p. 93)#

• Definitions. An antibiotic is produced by microorganisms and inhibits or kills others. Synthetic agents are “chemotherapeuticals”. “Antimicrobial” covers all of these.
• History. Salvarsan in 1910, sulphonamides in the 1930s. Fleming discovered penicillin (the chapter cites his 1929 paper and gives no discovery date), but it entered therapy only in 1941. “Shortly after”, it was used to treat animals.
• Stakes. “Antimicrobials are probably the single most important discovery in the history of medicine. They were considered miracle drugs” (p. 93). They have saved “millions of lives”.

9.1.1 Growth promotion (p. 93) - Accidental discovery. In the late 1940s, chickens fed tetracycline fermentation waste as a vitamin B12 source grew faster. The effect came from residual tetracycline (Stokstad and Jukes 1949). - Rapid embedding. “The practice of feeding sub-therapeutic doses of antimicrobials over long periods of time was readily adopted and soon became an integrated part of the production systems developed in industrialised animal husbandry” (p. 93). - Other observed effects at low doses (the chapter’s word; no magnitudes given): besides faster growth and/or better feed conversion, improved egg production, larger litters and higher milk yield. - Regulatory asymmetry. AGPs are “freely accessible and sold over the counter”. Therapeutic antimicrobials are mostly prescription-only. - Co-evolution with intensification. Four decades of AGP use “undoubtedly contributed to the development of current animal production systems”. AGPs “apparently” protect against diseases “promoted by intensification”. They allow “unphysiological early weaning and high stocking rates, raising questions on animal ethics and animal welfare” (p. 93).

9.1.2 Development of resistance (pp. 93–94) - Evolutionary framing. Resistance “follows the classical concept of survival of the fittest” and “has probably existed for as long as have bacteria” (p. 93). - Early general warning. Concern “more or less accompanied” penicillin’s introduction. Fleming (New York Times, 1945) warned misuse could mean “microbes are educated to resist penicillin”. By the early 1950s resistance “was well acknowledged in the medical, veterinary and pharmaceutical press” (p. 93). - So resistance was recognised at the very time growth promotion was discovered and adopted. My inference: the later dispute was less about whether resistance occurs than about whether low-dose feed use selected for it (Walton, p. 94), whether it could transfer (1977 Swedish view, p. 95), and whether any of this mattered for people. - Vertical to horizontal. Resistance was first thought to arise only by mutation and stay within the clone (vertical). “In the 1960s it was shown” that resistance genes on mobile elements can spread to other clones, species “and even to other genera” (horizontal) (p. 94).

9.2 The first early warning: the Swann Committee (p. 94)#

• Trigger and remit. In the mid-1960s, “growing concern over food-borne infections with multi-drug-resistant salmonella” led the UK to set up “an independent advisory committee in 1968”, chaired by Professor Michael Swann. Its remit was transferable resistance from growth-promoter and veterinary use. Its formal name was the Joint Committee on the use of Antibiotics in Animal Husbandry and Veterinary Medicine (p. 100).
• Evidence judgement. The data were “a sufficiently sound basis for action” (p. 94). Outside the section: the editors quote more of Swann, including that “the cry for more research should not be allowed to hold up our recommendations” (p. 181). That line is not in Ch. 9.
• Core recommendation. Non-prescription feed antibiotics should be restricted to those that (p. 94): 1. “are of economic value in livestock production under United Kingdom farming conditions”; 2. “have little or no application as therapeutic agents in man or animals”; 3. “will not impair the efficacy of a prescribed therapeutic antibiotic … through the development of resistant strains”.
• This is a gate combining a benefit test with two exclusion tests, not a ban. Criterion 2 depends on time. A class with little therapeutic use in 1969 could become critical later, and the chapter’s own examples show this happening (vancomycin, p. 94; Synercid, p. 96; everninomycin, p. 97). Criterion 3 would in principle catch such cases through cross-resistance, but only if the criteria were re-applied as human medicine changed. That re-application is what the standing committee and monitoring would have supplied (p. 94). My analysis; the chapter does not discuss this.
• Drug-specific recommendations were also made; the chapter gives one “for example”: “tylosin should not be available without prescription for use as a ‘feed’ antibiotic”.
• Institutional recommendation. A single permanent committee with “overall responsibility for the whole field of use of antibiotics and related substances whether in man, animals, food preservation, or for other purposes” (p. 94).

9.3 Subsequent action or inaction#

9.3.1 Implementation and dilution (pp. 94–95) - Reception. The recommendations “were based on less than full scientific certainty and created much debate and cries for more research, and faced strong opposition from the pharmaceutical industry and farming community” (p. 94). Even so, most were adopted in the UK “and later on” in the EU. - Dilution. “subsequent governments in the United Kingdom gradually diluted the recommendations”. The overarching committee “was not fully implemented” (so partly, not wholly, unrealised), and “no epidemiological studies to monitor antimicrobial resistance development were set up” (p. 94). No UK documents are cited on how or why. Outside the section, the editors put it more strongly: the recommendation “was not taken up, in the United Kingdom or elsewhere, for many years” (p. 174). - Decisions against Swann. - In 1975 the EU accepted the macrolides tylosin and spiramycin as growth promoters, which “has probably been one of the major reasons for the widespread macrolide resistance in, for example, enterococci and campylobacter from pigs” (p. 94). Hedged and uncited. - Avoparcin use was extended, e.g. to adult cattle, “against the Swann recommendations”, and grew from the mid-1970s, “at about the same time that its medical equivalent, vancomycin, started to come into hospital use” (p. 94). The chapter does not say who authorised the extension. - The dose argument. Low doses were argued to be “a special case”. Walton (1988): “In practical terms the use of a sub-lethal or a sub-inhibitory antibiotic concentration is therefore unable to select resistant strains from a bacterial population, and in this respect the Swann Report’s conclusion and recommendations were in error” (p. 94). - Rebuttal. “the recent bans on avoparcin, virginiamycin and tylosin followed the publication of studies demonstrating that this view (Walton’s amongst others) was wrong” (pp. 94–95). No citation is attached here. The chapter does not say whether the VRE and surveillance studies cited later (Klare, Bates, Aarestrup 1995; Aarestrup et al. 1998; DVL 1995; p. 96) are the ones meant. But DVL 1995’s reported finding that “the use of avoparcin as a growth promoter selects for vancomycin-resistant enterococci” (p. 96) bears directly on Walton’s practical claim that such use is “unable to select resistant strains”, even though the chapter does not connect the two passages. So the rebuttal is uncited where it is made, not unsupported within the chapter. - Global divergence. In the US, “low doses of tetracycline and penicillin are still used as feed additives for prophylaxis and growth promotion without veterinary prescription” (p. 95). This is a present-tense claim as of 2001.

9.3.2 The Swedish ban (pp. 95–96) - Background. “Similar to the situation in other countries, some Swedish scientists viewed the practice of routine addition of antimicrobials to animal feeds with scepticism”. Swann prompted “a broader debate” that led to a reassessment (p. 95). - 1977 official view. After Swann, a Swedish Board of Agriculture working group (LBS 1977) concluded that feed additives carry “a risk of increased resistance … but as the substances in use are mainly active against gram-positive bacteria from which resistance is not transferred, the impact of such development is negligible” (p. 95). It also noted “a negative attitude to all kinds of additives among consumers” and acknowledged benefits “in terms of increased production and prevention of certain diseases”. It proposed legislative changes, “especially in the requirements for approval”, “to mitigate possible risks” (p. 95). - So the eventual frontrunner’s own agricultural authority accepted a risk of more resistance but judged its impact negligible. The chapter reports this candidly. - Farmers move first. Farmers “were concerned that the continued use of antimicrobials might harm consumer confidence”. The Federation of Swedish Farmers (LRF) declared a policy of restricted use and in 1984 wrote to the Ministry asking for a ban (p. 95). The Ministry drafted the new Act “in response to the above”: the post-Swann debate, the 1977 reassessment and the farmers’ request. In the chapter’s account, no new scientific finding is cited as a trigger; on my reading, the farmers’ request, driven by consumer confidence, is the proximate one. (Table 9.2 gives the farmers’ motive as “health and consumer concerns”; the text mentions only consumer confidence.) - Legislation. The draft Feedingstuffs Act (Government Bill 1984/85) proposed restricting feed antimicrobials to “treatment, prevention or cure of diseases”, so that growth-promotion use “should not be allowed”. The grounds were: - risk of more resistance, “especially the risk of cross-resistance”; - “risk of increased susceptibility of animals to salmonella and other enteric pathogens” (a food-safety concern the chapter does not develop); - “uncertainty on the long-term effects of the continuous use of feed containing chemotherapeutics” (p. 95). - The Act passed in November 1985 and took effect in January 1986. Since then antimicrobials have been allowed only “for therapy and on veterinary prescription”. There is an unexplained gap between the Bill’s “prevention” and this “therapy”-only summary. - Outcome. “As a result”, total consumption was “greatly reduced from around 50 tonnes in 1985 to around 20 tonnes in 1996” (p. 95). The figure is sourced to SOU 1997, the first author’s commission. The context (“whether in feed or administered otherwise”) implies total animal use, but the chapter does not say so explicitly. This is the chapter’s only quantitative outcome. The causal link is not examined, and there are no data on animal health, productivity or costs after the ban. - EU accession. Sweden received a temporary derogation. “In support of the Swedish view”, the Ministry appointed a commission to review the science: the Edqvist commission, SOU 1997 (p. 95; p. 195). - Category finding. SOU 1997 found that “antimicrobial feed additives can at levels permitted in feeding stuffs, be used for treatment or prevention of animal disease, which is in violation of Council Directive 70/524/EEC” (p. 95). As amended (96/55/EEC, Art. 3a), that directive allows authorisation only if treatment or prevention is excluded at the permitted level. So, on SOU’s finding, the legal category “feed additive” did not match what the agents actually did. - Risk conclusion. “the risk for increased resistance associated with the general use of antibiotic growth promoters is far from negligible and the potential consequences are serious for both animal and human health” (p. 95). - Structure of the unknown. Magnitude is “difficult to fully establish because of the complexity of the problem and the lack of pertinent data”. (This sentence is unquoted and could be the authors’ voice or a paraphrase of SOU.) SOU set out the research needed as a 17-step causal chain. It put the minimum time at 5–10 years, and said research had to be done “for each resistance gene and for each antimicrobial substance, with subsequent updates of the risk assessment” (pp. 95–96). The chapter does not say whether 5–10 years is the total or applies to each gene and substance. - Who pays for waiting. It asked who “would bear the costs of waiting to do further research, or of taking action then to restrict antimicrobials — the risk-maker or the risk-taker?” (p. 96). The chapter does not report SOU’s answer. - Decision principle. The risk “can obviously not be excluded with certainty, nor can it be determined as acceptable” (p. 96). “Scientists may declare that the information is inadequate for decision making, but for the policymakers, failure to take action is not a neutral position but represents a positive decision to do nothing” (p. 96). “In a climate of uncertainty it is preferable to show caution” (p. 96). This passage reappears almost word for word, unquoted, in the chapter’s conclusions (p. 98).

9.3.3 The ban of avoparcin (p. 96) - Sentinel evidence. In March 1995, once “the first information on the occurrence of avoparcin- and vancomycin-resistant enterococci in pigs and poultry” was available (Klare 1995; Bates 1994; Aarestrup 1995), Danish farmers’ organisations agreed with the feed industry to stop using avoparcin voluntarily “to reduce the spread of antimicrobial resistance” (p. 96). The stated motive here is resistance, not consumer confidence, and the feed industry joined the halt rather than resisting it. - Government ban. Denmark banned avoparcin on 20 May 1995 under the safeguard clause of Directive 70/524/EEC, which lets a Member State temporarily suspend an additive on new information of danger to animal or human health. - Scientific basis (DVL 1995) was the demonstration of: - cross-resistance between avoparcin and vancomycin; - transferability of the resistance; - that “the use of avoparcin as a growth promoter selects for” VRE, and that VRE “can be transferred to humans via the food chain”. - These are pathway steps, not measures of human disease burden. - Diffusion. Norway suspended avoparcin in June 1995 and Germany banned it in January 1996. - Scientific adviser versus risk manager. In May 1996 SCAN concluded “further evidence was required to establish a risk”, while accepting “serious questions” had been raised. It said feed use should be reconsidered “at once if it were shown that transfer of resistance was possible from animal to human” (p. 96). - The Commission “however, proposed that in the climate of uncertainty and to avoid taking any risk, a temporary ban should be placed” EU-wide. The Standing Committee on Feedingstuffs agreed by qualified majority in December (1996, implied), and the ban took effect on 1 April 1997. - The chapter does not reconcile SCAN’s view that risk was not yet established with DVL’s “demonstration” that VRE “can be transferred to humans via the food chain”.

9.3.4 Danish ban of virginiamycin (p. 96) - On 16 January 1998, Denmark banned virginiamycin because of the risk of selecting streptogramin-resistant enterococci (Aarestrup et al. 1998). - The aim was “to protect human health and to preserve the lifespan of Synercid”, a streptogramin then in hospital trials and since licensed for multi-drug-resistant infections. This is a pre-emptive rationale: protecting a future human drug from resistance selected by a related farm product.

9.3.5 EU bans four AGPs (p. 96) - On 14 December 1998, EU agriculture ministers voted to ban virginiamycin, bacitracin zinc, tylosin phosphate and spiramycin from July 1999. The Commission presented this as “a precautionary measure to minimise the risk of development of resistant bacteria and to preserve the efficacy of certain antibiotics used in human medicine”. - Tylosin and spiramycin are the macrolides approved in 1975 against Swann: a 30-year gap from Swann’s tylosin recommendation. - Industry response. “The pharmaceutical industry protested against the decision and called for further scientific facts”. The virginiamycin manufacturer sought annulment “before the European Court of Justice”. A ruling was “not expected before the end of the year 2001” (p. 96). - The chapter is silent on what scientific advisers concluded about these four before the vote.

9.3.6 Avilamycin (p. 97) - Avilamycin, still EU-approved, showed cross-resistance with a potential human drug, everninomycin (Aarestrup 1998). The everninomycin manufacturers “have recently withdrawn it from hospital trial worldwide”. No reason is given, and the chapter draws no conclusion. The “However” suggests the withdrawal took away the immediate stake.

9.3.7 Scientific reports and recommendations (p. 97) - WHO 1997 (Berlin). “the magnitude of the medical and public health impact of antimicrobial use in food animal production is not known”. Yet its recommendations stated: “Increased concerns regarding risks to public health resulting from the use of antimicrobial growth promoters indicate that it is essential to have a systematic approach towards replacing growth promoting antimicrobials with safer non-antimicrobial alternatives.” - Copenhagen 1998. The EU chief medical officers’ conference, attended by medical and veterinary authorities, farmers, and the pharmaceutical and feed industries, reported: “Most of those at the conference considered the use of antimicrobials for growth promotion was not justified”. It called for alternatives “including better farming practice”. Note “most”: this was not a consensus. - SSC 1999. The Commission (DG XXIV) had asked the SSC to assess resistance “because of concern over … the rapidly increasing rate of development of antimicrobial resistance”. The EU Scientific Steering Committee called for “action … promptly to reduce the overall use of antimicrobials in a balanced way in all areas: human medicine, veterinary medicine, animal production and plant protection”. AGPs from classes which “are or may be used” in human or veterinary medicine “should be phased out as soon as possible and ultimately abolished”. - WHO 2000 Global Principles. AGPs in classes used, or submitted for approval, in humans “should be terminated or rapidly phased out in the absence of risk-based evaluations” (conditional wording).

9.4 Advantages and disadvantages (pp. 97–98)#

• Attention asymmetry. “Much less attention has been directed towards possible side-effects, and the number of independent scientific studies is relatively small compared to studies and congress contributions supported by the pharmaceutical industry” (p. 97). No counts are given.
• Trigger for independent research. It rose “significantly” only after the VRE–avoparcin link was found and “the possible consequences for human health were made clear” (p. 97).
• Evidence claim. “In the last few years substantial scientific evidence has shown that the use of antimicrobial growth promoters in food animals contributes to the problems of antimicrobial resistance in humans. This has most convincingly been shown for vancomycin-resistant enterococci” (p. 97). No citations are attached to this paragraph.
• Concession. Human medicine “undoubtedly is of more importance”, but “this cannot justify ignorance of potential human health risks” from animal use (pp. 97–98).
• Scale claim. Feed use is “one of the major sources of overuse and misuse of antimicrobials in animal farming” (p. 98). Not quantified.
• Breadth. The issue is “complex”: human and animal health, welfare, food safety, environment, production systems (p. 98). The text points to “Table 1”, printed as Table 9.1.

Table 9.1: Positive and negative effects of AGPs (p. 99; source: Edqvist and Pedersen)

• The table is the authors’ qualitative judgement, with no citations and no magnitudes.
• In several rows the benefit and the harm are two faces of the same effect. “Alleviates and dampens signs of disease” (welfare positive) is the other face of “camouflages stress associated with sub-clinical disease” (welfare negative) and “masks sub-clinical disease” (animal-health negative); “lower labour demand due to possibility of more intense production methods” is “intensification”. Other positives stand alone (better feed use, less manure, more production). So the table does not argue that all the benefits are harms, but for animal health, welfare and production systems it presents the benefit as the source of the harm.
• The “Human health: None” entry does not engage an argument made by AGP proponents, not in the chapter: that controlling enteric disease might reduce therapeutic use or pathogen load.
• The benefits go to production. The costs go to health systems, future patients and workers. This is the distributional core of the case, stated here rather than argued in the text.

9.5 Conclusions and lessons (p. 98)#

Covered in “The authors’ own lessons” below.

Table 9.2: early warnings and actions (p. 99; source: “EEA”) This table was compiled by the editors, not the authors. It has typos (“Flemming”, “tolysin”, “antimcirobials”) and several discrepancies with the text: - 1975 is garbled: “tolysin and spiramycin still permitted as growth promoters as human equivalents; vancomycin comes into use”. The text says the EU accepted them in 1975 against Swann (p. 94). - The SOU finding is narrowed to risk “in humans”; the text says consequences for “both animal and human health” (p. 95). - WHO 1997 becomes “essential to replace growth promoting antimicrobials”, dropping the finding that magnitude is “not known” (p. 97). - One manufacturer’s lawsuit becomes “Pharmaceutical industry … takes EU to the European Court” (p. 96). - The Danish 1995 and EU 1997 avoparcin bans are omitted, although they are the central regulatory events in the text. The Danish 1998 virginiamycin ban is also omitted. - The 1984 entry gives the Swedish farmers’ motive as “health and consumer concerns”; the text mentions only consumer confidence (p. 95). - WHO 2000 becomes a recommended “ban”; the text says “terminated or rapidly phased out”, and covers classes used “(or submitted for approval)” in humans (p. 97). - 1969 “severe restrictions” is the table’s own characterisation; the text describes a restrictive gate (p. 94). - 1977 says the Board “concludes it [the risk] is negligible”; the text says the working group accepted “a risk of increased resistance” and judged the impact negligible (p. 95). - 1985 dates the Swedish ban to the year the Act was passed (in force January 1986, p. 95), and gives its grounds as resistance “in animals”. The text does not limit the resistance risk to animals, and also cites cross-resistance and animals’ susceptibility to salmonella (p. 95). - Also absent from the table: the SCAN 1996 opinion and the Copenhagen 1998 conference (pp. 96–97).

References (pp. 99–100). There are 19: - 7 primary studies: Aarestrup ×3, Klare, Bates, Stokstad and Jukes, Fleming. - 3 reports or reviews: DVL 1995, Amabile-Cuevas and Chicurel, Walton. - 8 official documents: Swann, LBS, Government Bill, SOU, WHO ×2, Copenhagen, SSC. - 1 newspaper.

Two key official sources (SOU 1997, DVL 1995) come from the authors’ own institutions. The rebuttal of Walton (p. 95) and the “substantial scientific evidence” paragraph (p. 97) carry no specific citations. The Aarestrup et al. 1998 APMIS citation gives “Vol. 160”, which is probably Vol. 106 (to verify). Other probable citation slips (from general knowledge, to verify): the two Aarestrup papers’ journal “Microbiology Drug Research” is probably Microbial Drug Resistance; Stokstad and Jukes’s journal is probably Proceedings of the Society for Experimental Biology and Medicine; and Klare et al. lack Reissbrodt’s initial. SOU 1997’s “p. 132” is probably the report’s series number (SOU 1997:132), not a page (to verify).

Case timeline#

Lags between warning and action - Resistance knowledge vs adoption. Resistance was recognised (early 1950s) as growth promotion spread. The first animal-specific formal warning came some 15–20 years later (Swann 1969). - Swann → first national ban (Sweden 1986): 17 years. In the chapter’s account the proximate trigger was the farmers’ request over consumer confidence. The government’s stated grounds were resistance risks and long-term uncertainty, not a new finding (p. 95). - Swann → EU avoparcin ban (1997): 28 years. → four-AGP ban (1999): 30 years. For tylosin, Swann’s 1969 recommendation was followed by EU approval in 1975 and a ban effective in 1999. - Sentinel evidence (1994–95) → Danish ban: months to a year. → EU ban: about 2–3 years. At EU level the long lag lay before specific pathway evidence, not after it. Sweden is the exception: it banned in 1986 on general resistance risk and uncertainty, before that evidence existed (p. 95). - What was known when. The chapter is not fully consistent here. P. 94 dates to “the 1960s” the finding that resistance genes can spread to other clones, species “and even to other genera”. P. 98 presents as mainly 1990s findings that transferable resistance “is not restricted to certain bacteria (gram-negative)” and that genes move “even between genera”. The likely reconciliation, implied by p. 98’s parenthesis but not stated, is that 1960s knowledge concerned gram-negative bacteria and 1990s work extended it to gram-positives such as enterococci. The chapter does not itself connect the missing monitoring (p. 94) and scarce independent research (p. 97) to the two-decade wait for confirmation. The editors do (research “not followed up until the 1990s”, p. 172; “delayed”, p. 179; outside the section). - Harms and costs. None are quantified. Named harms are: macrolide resistance in pig enterococci and Campylobacter (“probably”, p. 94); VRE transferable via food (p. 96); streptogramin resistance (p. 96); erosion of new human drugs (pp. 96–97); and Table 9.1’s health-care costs, shorter economic life of drugs, occupational exposure, environmental gene pool and welfare. WHO 1997’s “not known” (p. 97) is the chapter’s own evidence that magnitude was unmeasured.

The authors’ own lessons and conclusions#

Lessons derived from the case evidence (p. 98)#

1. A competent mechanistic warning with little formal proof can be right. Swann’s warning “was based on a low level of scientific proof, but on a competent microbiological assessment that foresaw possible adverse consequences”. Its recommendations “were clearly precautionary, although the word ‘precaution’ was not actually used”. 1990s research confirmed it was “both accurate in its evaluation of data at the time and far-sighted in its assessment of future trends”. Grounded in the horizontal-transfer findings (pp. 94, 98) and the VRE pathway (p. 96).
2. The dilution was an epistemic error. Its justification rested “mainly” on “narrow considerations of what was precisely known rather than on taking account of what was not known, of the ignorance within the field”. Hence “science that embraces complexities, uncertainties and unknowns with more humility and less hubris is needed”. Grounded in the overturned Walton and 1977 Swedish reassurances (pp. 94–95). The UK decision-makers’ actual reasoning is not documented.
3. Non-regulatory stakeholders can act first. Farmers and their organisations “can take voluntary steps in advance of legislation to stop the use of products which cause concern and loss of confidence amongst consumers” (p. 98). Grounded in Sweden 1984 (p. 95) and Denmark 1995 (p. 96). The consumer-confidence motive is documented for Sweden only; the Danish halt is given a resistance motive.
4. Vindication. Farmers and Swann “have been vindicated by history”. This is a judgement based on mechanistic confirmation and later regulatory consensus, not on health-outcome data.

Recommendations and advocacy (normative, or beyond the chapter’s evidence)#

5. Committee composition. “scientific committees which are responsible for the evaluation of confidential information from industry, should consist of a broad panel of independent and up-to-date experts” from all implicated disciplines. “expertise from human medicine would have been particularly valuable for every risk assessment”. Support in the chapter: indirect only. No committee membership is described, and confidential industry data is first mentioned here.
6. Wider risk assessment. Include positive and negative impacts, “long-term microbiological and ecological effects”, and alternatives “such as better animal husbandry” (Table 9.1).
7. Decision rule. The first part repeats SOU 1997 almost word for word (three of the four sentences quoted on p. 96, without the “not a neutral position” sentence): risk that can be neither excluded nor judged acceptable calls for caution. The authors then add a clause of their own that is not in the SOU quotation: decision-making “needs to involve precaution, particularly when it is unacceptable, inhuman and unethical to wait for ultimate proof, when human fatalities could be involved”. It is normative, and in part it is the first author’s commission restating its own conclusion (pp. 96, 98, 195).
8. Common sense and foresight. “Common sense and far-sighted use of good scientific evidence which can predict serious impacts should not be ignored whilst waiting for ultimate proof.”
9. Implicit in the narrative, not listed: monitor resistance (p. 94); fund independent research (p. 97); govern antimicrobial use across all sectors (Swann’s committee, p. 94; the SSC’s “all areas”, p. 97).

Mechanisms and dynamics#

1. Accidental discovery, fast adoption and co-evolution (lock-in). - The growth effect was a by-product (p. 93). It was adopted “readily”, with no pre-adoption appraisal of resistance mentioned, and became “integrated” into industrial husbandry. - It let the system change shape (early weaning, high stocking, p. 93; intensification, Table 9.1), so the new shape came to depend on it. - Withdrawal then looks costly because the system has adapted around the input. The chapter asserts this more than it shows it. Sweden’s ban (p. 95) shows withdrawal was possible, but the chapter gives no data on what it cost, and it does not itself present Sweden as proof of reversibility.

2. Classification by purpose, not function. - The same pharmacology sat in two regimes: over-the-counter feed additive versus prescription drug (pp. 93, 95). - The feed-additive directive (as amended, Art. 3a) made authorisation conditional on treatment or prevention of animal disease being excluded at the permitted level. The Swedish commission found that AGPs at permitted levels could treat or prevent disease (p. 95). - Assessment fell to feed and animal-nutrition bodies (SCAN, p. 96). - So the label, set by stated purpose and dose, decided which gatekeepers, evidence standards and expertise applied. The committee-composition lesson (p. 98) is the remedy.

3. Reassurance anchored on the current edge of knowledge. The chapter documents three successive confinement assumptions, each later overturned: - (a) resistance confined to the clone, overturned in the 1960s (p. 94); - (b) resistance “not transferred” from gram-positive bacteria (1977, p. 95), overturned in the 1990s (p. 98); - (c) sub-inhibitory doses “unable to select resistant strains” (Walton 1988, p. 94), said to be disproved (pp. 94–95; uncited there, though DVL 1995’s finding that growth-promoter use of avoparcin selects VRE, p. 96, contradicts it). - The chapter presents each as a scientific position, not as bad faith; it says nothing about motives. Only (b) and (c) are shown being used to reassure about farm use; (a) is reported as a general scientific belief (“Originally it was thought”, p. 94). My interpretation: each turned a gap in knowledge into a claim of safety, and the reassurance shifted to new ground as each was overturned. The chapter’s own diagnosis is weight on “what was precisely known” over “what was not known” (p. 98).

4. Burden of proof; calls for research that in effect delay. - Swann acted on “less than full scientific certainty” and drew “cries for more research” (p. 94). - SCAN required “further evidence … to establish a risk” (p. 96). The burden was on showing harm. - Industry “called for further scientific facts” and then litigated (p. 96). - SOU set out that the research needed to establish the risk would follow a 17-step causal chain, take at least 5–10 years, and have to be done per gene and per substance (pp. 95–96). - Where causation is long, multi-step and variant-specific, “prove harm first” is slow to satisfy. SOU put the minimum at 5–10 years, with further updates as genes and substances multiply. So demands for research in practice preserve the status quo for years (my inference from pp. 95–96). The chapter documents the calls and their timing but does not allege bad faith. SOU’s reframing, that inaction “is not a neutral position” (p. 96), turns delay into a decision with its own distribution of costs (“risk-maker or the risk-taker?”).

5. Asymmetric knowledge production and missing monitoring. - Research went to benefits. Independent side-effect studies were few compared with industry-supported studies and “congress contributions” (p. 97). - Independent work rose only after a salient signal (p. 97). - Swann’s monitoring was never set up (p. 94). - So the evidence gap was partly made by institutions (my inference). Ch. 9 does not show the gap being cited as reassurance. Outside the section, the editors make the “no evidence of harm” versus “evidence of no harm” point explicitly for BSE. They then cite antimicrobials only as a case of “similar delays in the conduct of relevant scientific research” (p. 172).

6. Fragmentation and gradual dilution. Swann’s cross-sectoral committee “was not fully implemented” (p. 94). Dilution came “gradually” through “subsequent governments”, product approvals (1975) and extensions of use (avoparcin) (p. 94). A warning can be formally accepted and still eroded through small exceptions and missing oversight machinery. The chapter shows this through outcomes, not decision records.

7. Scientific adviser versus risk manager. SCAN (1996) said risk was unproven; the Commission acted “to avoid taking any risk” (p. 96). The 1998 ban was framed as “precautionary”, and a supplier then went to court (p. 96). Under uncertainty, the contest moves from “what does science say?” to “who decides, on what standard?”. The chapter backs the risk manager but does not examine legitimacy. “To avoid taking any risk” is a near-zero-risk formulation that the chapter does not question.

8. Frontrunners and multi-level law. - Sweden banned unilaterally and then needed an accession derogation (p. 95). - Denmark used the EU safeguard clause (p. 96). - Norway and Germany followed, then the EU (p. 96). - National flexibilities let early movers act and produce domestic evidence (DVL 1995, SOU 1997), which then pressed the collective regime. The harmonised regime also constrained them. The editors call EU membership “an institutional obstacle” for Sweden (p. 180, outside the section).

9. Split interests along the value chain; trust as a driver. - In the UK, pharma and farmers opposed Swann (p. 94). - In Sweden, farmers broke ranks over “consumer confidence” (p. 95). - In Denmark, farmers’ organisations and the feed industry together stopped avoparcin voluntarily, with the stated aim “to reduce the spread of antimicrobial resistance” (p. 96). No consumer-trust motive is given there. - Pharma, as supplier, kept opposing and litigating (p. 96). - Users whose market rests on public trust in the end product can drop an input when its reputational risk outweighs its productivity gain. The Swedish case shows this directly; the Danish case shows users and an intermediate supplier (feed makers) acting on a resistance rationale. Consumer attitude (“a negative attitude to all kinds of additives”, p. 95) was a factor before the science was settled. The chapter treats this as legitimate. The editors go further, claiming that heeding “public misgivings” would have held resistance “in check” (p. 178, outside the section; beyond Ch. 9’s evidence).

10. Distribution: a shared, depletable resource. - Benefits go to producers (p. 93; Table 9.1). Costs go to health systems, future patients and workers (Table 9.1). - Three drug pairs show farm agents selecting cross-resistance to human drugs: avoparcin/vancomycin (p. 94), virginiamycin/Synercid (p. 96), avilamycin/everninomycin (p. 97). Two involved drugs still in trials. - Efficacy behaves like a common-pool resource. Use in one sector draws down a stock another depends on, including drugs not yet marketed. The language of “lifespan” (p. 96) and “economic life” (p. 99) points to this, but the chapter does not develop the economics or quantify anything.

11. The “minor contributor” defence. Human medicine matters more (pp. 97–98), but that “cannot justify ignorance” of animal use. The SSC called for reductions “in all areas” (p. 97). With many sources of a harm, each can point to a bigger one. The chapter rejects that move without quantifying the animal share.

12. Masking. Table 9.1 lists: “masks sub-clinical disease”, “camouflages stress”, “camouflages bad feed quality”, “limits incentives for hygienic improvements”, “hampers … development of alternatives” (p. 99). A technology that suppresses the symptoms of a system’s problems both reduces pressure to fix them and slows alternatives. This deepens lock-in (§1). It is asserted, not evidenced.

13. How the actors thought. - Proponents and some scientists: safety came from confinement (§3). “Sub-therapeutic” dosing did double work, as a scientific argument and as the basis of the feed-additive category (§2). Confidence was high: Walton called Swann “in error”. - Industry: industry-supported studies and congress contributions outnumbered independent ones, and side effects got “much less attention” (p. 97). The chapter names no actor for that neglect. Industry then asked for facts and litigated (p. 96). - Regulators before 1995: partial acceptance, then exceptions (p. 94). SCAN’s model was “prove transfer, then reconsider” (p. 96). - Precautionary actors (Swann, SOU, DVL, Commission): plausible mechanism plus serious consequences justified action. Inaction is a decision. - Farmers: market standing and consumer trust first (p. 95). The editors add that Swedish farmers had knowledge of alternative husbandry (p. 178). Ch. 9 only gestures at “better animal husbandry” (p. 98).

14. Framing and language. - “Miracle drugs” (p. 93) sets the stakes. - “Growth promoter” names the product by its production benefit, not its pharmacology. This is my interpretation; the chapter does not comment. - “Sub-therapeutic” and “low dose” (pp. 93–94) marked the use off from medicine, scientifically and legally. - “Survival of the fittest” and “educated to resist” (p. 93) make resistance predictable, supporting the claim that it was foreseeable. - “Cries for more research” (p. 94) is pejorative. It echoes Swann’s own warning, quoted by the editors, that “the cry for more research should not be allowed to hold up our recommendations” (p. 181, outside the section). - “Climate of uncertainty”, “to avoid taking any risk” and “not a neutral position” (p. 96) reframe inaction as a choice. - “Humility … hubris”, “vindicated by history”, “common sense” and “ultimate proof” (p. 98) are rhetorical. “Ultimate proof” risks being a straw standard: the actors shown asked for “further evidence” (SCAN) and “further scientific facts” (industry), not ultimate proof (p. 96).

15. Complexity, time and irreversibility. - SOU cites “the complexity of the problem” (p. 95). Genes move across genera (p. 98) and build an “environmental pool” (Table 9.1). Effects are distributed, delayed and hard to attribute. - Irreversibility is implicit, not argued, in Ch. 9. The editors state it (“Resistance to antimicrobials, it is now known, is long term”, p. 171, outside the section). Ch. 9 has no data on whether resistance falls after withdrawal.

16. Innovation and alternatives. The chapter quotes WHO 1997 and the Copenhagen conference calling for “safer non-antimicrobial alternatives”, Copenhagen adding “including better farming practice” (p. 97). In their own voice, the authors call for assessments to weigh “alternative options, such as better animal husbandry” (p. 98), and their Table 9.1 says AGPs hamper alternatives. It gives no evidence on how alternatives performed, what they cost or how widely they were taken up; the Swedish tonnage is the only datum. The editors’ claim that curtailment fostered innovation (p. 182, outside the section) is not supported by Ch. 9.

Transferable insights (technology-neutral)#

1. Regulating by stated purpose rather than actual function can put an active agent outside the controls applied to the same agent in another use. The label then decides which experts, standards and gatekeepers apply. - Evidence: pp. 93, 95 (over-the-counter vs prescription; SOU finding against the directive’s own conditions); p. 96 (animal-nutrition committee). - Moderate. The regulatory facts are clear; their contribution to harm is inferred, not measured.

2. Safety reassurances can rest on a mechanistic model that treats the current edge of knowledge as a boundary of risk. As knowledge moves, the reassurance shifts to new ground. - Evidence: the gram-positive (1977; pp. 95, 98) and dose (Walton 1988; pp. 94–95) assumptions, both later overturned; plus the earlier clone-confinement belief (p. 94), which the chapter reports as general science rather than as a reassurance about farm use. - Moderate. There are two documented reassurances within this case, and the rebuttal of the dose argument is uncited where it is made (DVL 1995, p. 96, supports it without being linked to it). The “retreat” pattern is my reading, not the chapter’s.

3. Where proving harm needs a long causal chain re-established for every variant, “prove harm first” in effect means continuing for years, and the question reopens with each new variant. Inaction under uncertainty is a decision with its own distribution of costs. - Evidence: pp. 95–96 (17 steps; minimum 5–10 years; research per gene and substance with later updates; “not a neutral position”; “risk-maker or the risk-taker?”). - Moderate. The evidentiary structure is well described. The normative conclusion comes from an interested source, and no counterfactual is shown.

4. Evidence that would test a warning is rarely produced unless someone is mandated and funded to produce it. When monitoring is never set up, the result is an absence of evidence. - Evidence: p. 94 (no monitoring); p. 97 (few independent studies until the VRE signal). - Moderate for the gap itself: the non-implementation is clearly stated, though the research imbalance is unquantified. Suggestive for any claim that the gap was then used as reassurance, which Ch. 9 does not document (the editors make that point for BSE, p. 172).

5. Research effort follows interest. Benefits are studied by those who gain; side effects stay under-studied until a salient event forces attention. - Evidence: p. 97. - Suggestive. Plausible but asserted without counts or references.

6. A well-reasoned early warning based on mechanism can be substantially right decades before confirmation. Such warnings can be eroded through incremental exceptions and missing infrastructure rather than open repeal. - Evidence: pp. 94, 98. - Moderate for this case. The confirmation is mechanistic, not outcome-based. The dilution is shown through outcomes, not decision records. It is one case, so “tend to” would overreach.

7. Adopting a warning’s substantive rules without its coordinating and oversight machinery is fragile implementation. - Evidence: p. 94 (cross-sectoral committee “not fully implemented”; no monitoring; dilution followed). - Suggestive. One case; the causal link is implied.

8. If a hazard’s main harms fall outside the assessing body’s discipline, assessment framed by the producing sector’s discipline can sideline them. Committee composition shapes what counts as risk. - Evidence: p. 98 (human medicine expertise); p. 96 (SCAN). - Asserted/suggestive. Committee membership is never documented. The editors make the point more explicitly (p. 174).

9. Downstream users may break with suppliers when their own market depends on public trust. Voluntary user action can precede and enable regulation. - Evidence: p. 95 (Swedish farmers, 1984, consumer confidence); p. 96 (Danish farmers with the feed industry, March 1995, “to reduce the spread of antimicrobial resistance”, followed by a government ban within about two months). - Moderate for the second sentence: two documented instances of user action preceding regulation. Suggestive for the trust motive, which is documented only for Sweden. Both are Nordic cases, which limits generalisation; the chapter says nothing about the size or market position of either sector.

10. A technology that props up a production system can mask that system’s problems and weaken incentives for alternatives, deepening dependence.

    • Evidence: p. 93; Table 9.1, p. 99.
    • Asserted. A coherent argument, from the authors’ own uncited table.

11. When one use of a shared, depletable resource degrades it for others, including future uses not yet deployed, the costs fall outside the decision. Early deployment in one domain can pre-emptively erode options in another.

    • Evidence: pp. 94, 96, 97 (three cross-resistance pairs); p. 99 (“shortens economic life”).
    • Moderate. The pairs are cited to primary studies; the impact is unquantified.

12. Showing that another source dominates a harm does not justify ignoring a contributing source.

    • Evidence: pp. 97–98; SSC “all areas”, p. 97.
    • Asserted. A normative argument, fairly made; the animal share is unquantified.

13. Early-mover jurisdictions using legal flexibilities can generate the evidence and demonstration that later drives collective action, while the collective regime constrains them.

    • Evidence: pp. 95–96 (Swedish derogation; Danish safeguard ban; Norway → Germany → EU; Danish virginiamycin → EU four-AGP ban).
    • Moderate. The sequence is well documented; causation is implied by the sequence.

14. Once a concrete harm pathway (a sentinel finding) is documented, action can come within months, after decades of inaction on a general warning. The long lag lies before specific evidence.

    • Evidence: p. 96 (VRE 1994–95 → Danish ban May 1995 → EU April 1997) against 26–28 years from Swann.
    • Strong for the timeline in this case; moderate as a generalisation. Caveats: Sweden (1986) acted without such evidence (p. 95), and the EU acted even though SCAN judged the pathway evidence insufficient (p. 96). So the sentinel finding enabled action rather than settling the science.

15. Uncertainty about magnitude is not uncertainty about direction or mechanism. Expert bodies can coherently recommend action while stating the size of harm is unknown.

    • Evidence: p. 97 (WHO 1997); p. 95 (SOU); p. 96 (Commission despite SCAN).
    • Moderate. Several bodies took this stance, WHO most clearly independently of the authors (SOU was Edqvist’s commission). Whether each action was justified is separate.

16. When scientific advisers and decision-makers diverge under uncertainty, the dispute shifts to legitimacy, legal standard and forum. Litigation by affected suppliers is a predictable next step.

    • Evidence: p. 96.
    • Suggestive. One sequence, with its outcome unknown when the chapter was written.

Limitations, contestation and bias check#

Standpoint and self-citation. - Edqvist chaired the SOU commission (p. 195), and Pedersen directed DVL (p. 198). Neither link is disclosed in the chapter; both appear only in the author annex, under a general disclaimer that chapter views are the authors’ own (p. 195, n. 14). - The concluding decision principle (p. 98) reuses the SOU text (p. 96), and then adds the authors’ own “inhuman and unethical” clause. - The SOU commission was, in the chapter’s own words, appointed “in support of the Swedish view” (p. 95). - None of this is improper: these were leading expert bodies. But the chapter is partly advocacy for the authors’ institutions’ past positions, and its “lessons” partly restate a Swedish policy doctrine.

Evidence of harm is mechanistic, not outcome-based. - There are no figures for illness, deaths or costs. The chapter’s own WHO source says magnitude “is not known” (p. 97). - “Substantial scientific evidence” (p. 97) carries no citation. It appears to rest on the 1994–95 VRE studies and DVL 1995 cited on p. 96. - “Vindicated by history” (p. 98) rests on confirmed mechanisms and regulatory convergence, not demonstrated health outcomes. That is consistent with a precautionary argument, but it is not proof that dilution caused a measured burden. - The macrolide claim is hedged and uncited (p. 94). The Swedish tonnage drop is causal “as a result” without examination (p. 95).

Counter-arguments thin or absent. - The scientific case that growth-promoter use was safe appears only through Walton, whose affiliation the chapter does not give. The rebuttal is uncited where it is made (pp. 94–95), though DVL 1995’s selection finding (p. 96) supports it. No industry-sponsored study is cited or described, although p. 97 says such studies outnumbered independent ones. - SCAN’s reasoning is summarised in two sentences. Its disagreement with DVL is left unexplained (p. 96). - Scientific advice before the 1998 four-AGP vote is not reported. From general knowledge (to verify), SCAN reportedly did not back an immediate virginiamycin ban in 1998. If so, the Council overrode its scientific committee a second time, and the chapter leaves this out. - There are no data on the post-ban animal health, productivity, welfare or economics in Sweden, or on whether therapeutic use rose. Critics later made such claims (see Open questions). - Table 9.1’s “Human health: None” is not defended. - Benefits are unquantified, so the chapter’s own call for weighing “both positive and negative impacts” (p. 98) cannot be carried out from its contents.

Hindsight and narrative selection. - Casting Swann as “the first early warning” and as “clearly precautionary” though “the word … was not actually used” (p. 98) is retrospective labelling. It is fair, but it builds the frame into the history. - Swann’s time-dependent criterion 2 is not critiqued, although the chapter’s own evidence shows therapeutic relevance changing (pp. 94, 96, 97). Criterion 3 would have covered this only with ongoing re-assessment. - To its credit, the chapter shows the frontrunner’s own 1977 authority accepting the reassuring view, and that the Swedish ban came from farmers and uncertainty rather than new science (p. 95). This shows that even the “good” actors initially shared the dominant model. - “Ultimate proof” (p. 98) risks being a straw standard; the opponents shown asked for “further evidence” and “further scientific facts” (p. 96).

Case selection and fairness. - This is a partial success case with Nordic early movers. It shows well that precaution can work, but it is weak on precaution’s costs. - In the other direction, the chapter: - acknowledges benefits (p. 93; Table 9.1); - concedes that human medicine is the bigger driver (pp. 97–98); - is honest about uncertainty (pp. 95–96, 98); - hedges causal claims (“probably”, “apparently”). - Independent bodies (WHO 1997, 2000; Copenhagen 1998; SSC 1999) reached broadly similar conclusions (p. 97), so the position was mainstream in Europe by 2001.

Internal inconsistencies. - Table 9.2 has discrepancies (see above). - The text points to “Table 1” for Table 9.1. The reference list has a likely APMIS volume error and probable journal-title slips (see References). (The split “B / 12” in the extract, p. 93, is a text-extraction artefact of the vitamin’s sub/superscript, not a typo in the chapter.) - Dating of horizontal transfer. P. 94 dates spread “even to other genera” to the 1960s. P. 98 presents gene movement “even between genera” as mainly a 1990s finding. The chapter does not reconcile the two (see “What was known when”). - Swedish Act scope. The Bill allowed use for “treatment, prevention or cure”, but the chapter says that since 1986 use has been allowed only “for therapy and on veterinary prescription” (p. 95). - Outside the section, the editors date Swann to “1967” (p. 172) and to a “1968 recommendation, by the Swan Committee” (p. 174), against the chapter’s 1968 set-up and 1969 report (p. 94).

How the editors amplify the case (outside the section; cross-reference only). The synthesis chapter cites Ch. 9 often and sometimes goes beyond it: - “humility … hubris” (p. 170); - sensitivity to ignorance “overwhelmed by scientific over-confidence” (p. 171); - concerns “not followed up until the 1990s” (p. 172), grouped with BSE’s “no evidence of harm” pattern, though for antimicrobials only the research delay is claimed; - a disciplinary “blind spot”, with veterinary framing sidelining human impacts (pp. 173–174); - a counterfactual: had Swann’s recommendations “been implemented and sustained, they might at least have mitigated the difficulties now recognised” (p. 173); - “as the antimicrobials case study illustrates”, appraisals should specify the research question, the time and funding it needs, and the independence of whoever carries it out; they should also say, “as the Swann Report did”, whether hazard-reducing action should come before or after the research (p. 173); - Swann’s 1968 evidence “allowed explicit anticipation of a series of potential environmental, animal welfare and human health problems” (p. 173), which goes beyond Ch. 9, where Swann is credited with foreseeing resistance consequences only (p. 98); - “poor practice in the agricultural use of antimicrobials” as deliberate non-compliance (p. 175), which Ch. 9 does not document; - Sweden’s reduction as an example of “wider assessment of the pros and cons” (p. 176); - husbandry alternatives “not actively promoted” (p. 177); - farmers’ lay knowledge, and heeding “public misgivings” would have held resistance “in check” (p. 178), which is beyond Ch. 9; - critical research “delayed” (p. 179); - Swann’s “cry for more research should not be allowed to hold up our recommendations” (p. 181); - curtailment fostering innovation (p. 182), with no evidence in Ch. 9; - the glossary’s example of uncertainty (“known” impacts, “unknown” probabilities), antibiotics in feed 1969–present (p. 192).

Anyone citing “the antimicrobials lesson” should separate what Ch. 9 shows from what the synthesis attributes to it.

Notable quotes#

1. “The practice of feeding sub-therapeutic doses of antimicrobials over long periods of time was readily adopted and soon became an integrated part of the production systems developed in industrialised animal husbandry.” (p. 93)
2. “microbes are educated to resist penicillin” (Fleming 1945, quoted p. 93)
3. The data were “a sufficiently sound basis for action” (Swann 1969, quoted p. 94)
4. “In practical terms the use of a sub-lethal or a sub-inhibitory antibiotic concentration is therefore unable to select resistant strains from a bacterial population, and in this respect the Swann Report’s conclusion and recommendations were in error.” (Walton 1988, quoted p. 94)
5. “Scientists may declare that the information is inadequate for decision making, but for the policymakers, failure to take action is not a neutral position but represents a positive decision to do nothing.” (SOU 1997, quoted p. 96)
6. “the risk-maker or the risk-taker?” (p. 96)
7. “Much less attention has been directed towards possible side-effects, and the number of independent scientific studies is relatively small compared to studies and congress contributions supported by the pharmaceutical industry.” (p. 97)
8. “narrow considerations of what was precisely known rather than on taking account of what was not known, of the ignorance within the field” (p. 98)
9. “science that embraces complexities, uncertainties and unknowns with more humility and less hubris is needed.” (p. 98)
10. “Common sense and far-sighted use of good scientific evidence which can predict serious impacts should not be ignored whilst waiting for ultimate proof.” (p. 98)

Open questions#

Within the text 1. Why did SCAN (1996) judge that further evidence was needed to establish a risk when DVL (1995) said food-chain transfer was demonstrated (p. 96)? What did SCAN conclude on the four AGPs before the December 1998 vote? 2. What were the UK “dilution” decisions (p. 94)? Who made them, on what reasoning, and with what industry input? 3. Which studies refuted the sub-inhibitory dose argument (pp. 94–95), and when? 4. What happened to Swedish animal health, productivity, therapeutic use and farm incomes after 1986 (p. 95)? 5. What share of EU antimicrobial use in the 1990s was growth promotion (p. 98)? 6. Why did avilamycin remain approved after 1999 (p. 97)? 7. What confidential industry data did committees assess, and did confidentiality limit scrutiny (p. 98)? 8. Did any regime re-evaluate a feed antibiotic when a related class entered human medicine (Swann’s criterion 2)?

Pointers for the hindsight strand (from my general background knowledge, not from the report, and not verified in this session; check against primary sources) - Litigation. My understanding is that the challenge was Pfizer Animal Health v Council (T‑13/99), decided by the Court of First Instance on 11 September 2002, which dismissed the action with extended discussion of the precautionary principle. A parallel bacitracin-zinc case (Alpharma, T‑70/99) was decided the same day. If so, “European Court of Justice” (p. 96) is imprecise. - EU. A full ban on antibiotic growth promoters is generally dated to 1 January 2006 (Reg. (EC) 1831/2003). - US. The FDA’s 1977 proposal to withdraw feed penicillin and tetracyclines is reported to have stalled amid congressional calls for more research. Growth-promotion uses of medically important antimicrobials reportedly ended in January 2017 (FDA Guidance #209/#213). This is directly relevant to p. 95 and to Mechanism §4. - Post-ban evidence (contested). DANMAP data are widely cited as showing large falls in VRE and in streptogramin- and macrolide-resistant enterococci in Danish food animals after the bans. A WHO review of Denmark’s AGP termination (2002/03) is reported to have found little overall productivity loss, but more weaner-pig diarrhoea and higher therapeutic use in that group. Critics (e.g. Casewell et al. 2003; Phillips et al. 2004, J Antimicrob Chemother) argued the bans raised therapeutic use and animal disease with no demonstrated human benefit. European hospital VRE reportedly rose in the 2000s through hospital-adapted lineages, which complicates the attribution on p. 97. - Dose argument. Later work (e.g. Gullberg et al. 2011, PLoS Pathogens) is reported to show selection far below inhibitory concentrations, bearing directly on Walton (p. 94). - Recurrence. Transferable colistin resistance (mcr-1), reported from food animals in 2015–16, is widely cited as a repeat of the same pathway outside Europe. - Sweden. Check SVARM data on the 50 t → 20 t figure. Check also whether ionophore coccidiostats remained permitted in feed after 1986. If so, “only … for therapy and on veterinary prescription” (p. 95) needs a caveat.

Audit log#

Independent audit, 2026-09-25. Checked against the full text extract (pp. 93–100), a layout text extraction of p. 99 (Tables 9.1 and 9.2; page images could not be rendered in this session), and the cited out-of-section pages (PDF p. 6; pp. 170–182, 192, 195, 198). All direct quotes from Ch. 9 checked verbatim; the numbers, dates and names in the section notes and timeline were confirmed except where changed below.

1. Standpoint: added the author-annex disclaimer (p. 195, n. 14) and corrected the timing of the commission’s appointment relative to the accession derogation.
2. Panels/voices: added the out-of-section note that Jim Bridges (another chapter’s author) sat on SCAN in 1991–97 and on the SSC from 1997 (p. 195).
3. Absent voices: removed the implicit attribution of Walton to industry (the chapter gives no affiliation), and narrowed “no UK government sources” to exclude the Swann Report itself.
4. 9.1: Fleming “discovered penicillin in 1929” corrected (the chapter cites the 1929 paper and gives no discovery date).
5. 9.1.2: softened “the open question was never whether resistance occurs” to a labelled inference that recognises Walton’s dispute over selection.
6. 9.2: rewrote the criterion-2 critique to note that criterion 3 would catch later cross-resistance if re-applied, and labelled it as my analysis.
7. 9.3.1: the committee was “not fully implemented” (not unbuilt); added the editors’ stronger p. 174 wording, flagged as outside the section.
8. 9.3.1: the avoparcin extension is no longer attributed to the EU (the chapter names no actor); retitled “EU decisions” to “Decisions against Swann”.
9. 9.3.1: nuanced “studies not cited” (no citation at that point; unclear whether the p. 96 studies are the ones meant).
10. 9.3.2: added the omitted “some Swedish scientists viewed … with scepticism” background (p. 95).
11. 9.3.2: 1977 view corrected to “risk exists but impact negligible” (also in the timeline).
12. 9.3.2: ban trigger nuanced (“in response to the above”); flagged Table 9.2’s addition of “health” to the farmers’ motive.
13. 9.3.2: added the Bill’s “prevention” versus the “therapy only” tension.
14. 9.3.2: SOU’s “5–10 years” clarified as a minimum, with the per-gene/per-substance scope noted as ambiguous; flagged the unquoted “magnitude” sentence as possibly the authors’ voice.
15. 9.3.3: added the Danish voluntary halt’s stated motive (“to reduce the spread of antimicrobial resistance”) and the feed industry’s joint role.
16. 9.3.3: replaced the non-verbatim SCAN “not shown” quotation with an accurate paraphrase (also in Open question 1).
17. 9.3.6–9.3.7: added the “However” reading for everninomycin, the SSC’s remit, and the SSC’s “are or may be used” wording.
18. 9.4: corrected the “made human consequences clear” paraphrase to “possible consequences for human health were made clear”.
19. Table 9.1: softened “the table argues the benefits are the harms” (several positives stand alone); labelled the enteric-disease counter-argument as proponents’, not the chapter’s.
20. Table 9.2: added four further discrepancies (Danish virginiamycin ban omitted; the 1984 “health” motive; WHO 2000 “ban”; 1969 “severe”).
21. References: added probable journal-title and author-initial slips (to verify).
22. Timeline: 1949 changed to “late 1940s”; Salmonella trigger changed to “growing concern”; 1970s date sourced to Table 9.2; derogation date changed to c. 1994–95 (not given in the chapter); Danish motive added.
23. Lags: “about 20 years” changed to “15–20 years”; the Swedish ban trigger restated; EU lag given as 2–3 years; Sweden noted as the exception to “lag before pathway evidence”.
24. What was known when: flagged the p. 94 versus p. 98 inconsistency on cross-genera transfer; removed the claim that the chapter “implies” monitoring gaps explain the delay (the editors say so, pp. 172, 179).
25. Authors’ lesson 3: noted the consumer-confidence motive is documented only for Sweden.
26. Authors’ lesson 7: separated the SOU sentences reused verbatim from the authors’ own “inhuman and unethical” clause.
27. Mechanism 1: removed the claim that the chapter offers Sweden as proof of reversibility.
28. Mechanism 3: removed the unsourced “none was fraudulent / sincere” assertions; noted that (a) is general science, not a documented reassurance; labelled “retreat” as my interpretation.
29. Mechanism 4: heading changed from “means of delay” to “in effect delay”; replaced “cannot be met within any relevant timescale” with SOU’s actual 5–10-year minimum.
30. Mechanism 5: corrected the p. 172 cross-reference (the editors’ “no evidence of harm” point is about BSE; for antimicrobials they claim only research delay).
31. Mechanism 6: “left unbuilt” changed to “not fully implemented”.
32. Mechanism 9: corrected the Danish motive and actors; the trust mechanism is now tied to Sweden only.
33. Mechanism 13: removed the attribution to industry of leaving side effects under-studied (the chapter names no actor).
34. Mechanism 14: noted that “cries for more research” echoes Swann’s own phrase (p. 181); “straw standard” softened and both opponent formulations cited.
35. Mechanism 16: corrected a misattribution. “Safer non-antimicrobial alternatives including better farming practice” is WHO 1997/Copenhagen 1998, quoted in the chapter, not the authors’ own call; the authors’ own wording (p. 98) is now given.
36. Mechanism 8: page reference for Norway/Germany/EU corrected to p. 96.
37. Insight 2: downgraded from Strong to Moderate (two documented reassurances; dose rebuttal uncited); removed “sincere”.
38. Insight 3: “continuing indefinitely” changed to “continuing for years, reopened with each new variant”.
39. Insight 4: split the strength rating (Moderate for the gap; Suggestive for its use as reassurance).
40. Insight 6: “tend to” changed to “can”; noted it is a single case.
41. Insight 7: “committee not built” changed to “not fully implemented”.
42. Insight 9: corrected the Danish evidence; split the strength rating (Suggestive for the trust motive).
43. Insight 14: added the caveats that Sweden acted without pathway evidence and the EU acted despite SCAN.
44. Insight 15: noted that SOU is not independent of the authors.
45. Limitations: added the n. 14 disclaimer and the authors’ added clause; “substantial evidence” now “appears to rest” on the p. 96 studies; Walton de-attributed from industry; criterion-3 caveat added.
46. Internal inconsistencies: removed the “B12 superscript typo” (an extraction artefact); added the cross-genera dating and Bill-scope inconsistencies.
47. Editors’ amplification: added the p. 173 counterfactual and the p. 176 “wider assessment” citation; qualified the p. 172 item.
48. Digest: corrected the Edqvist commission dates, “In 1949”, “same drug classes”, the Swann gate, “no cross-sectoral committee”, “risk negligible”, the Danish motive, the SCAN concession, and “5–10 years per gene”; added the authors’ extra clause to the decision rule; Insight 2 downgraded and Insights 3/4/6/7/10 qualified; Walton de-attributed; internal wrinkles and further Table 9.2 discrepancies added.

Second-pass audit, 2026-09-25 (resumed session). Re-read the full extract (pp. 93–100). Rendered all eight chapter pages as images: Tables 9.1 and 9.2 match the notes, the table sources are confirmed (9.1: Edqvist and Pedersen; 9.2: EEA), and there are no boxes or figures. Re-checked the out-of-section citations against PDF text (p. 6; pp. 170–182, 192, 195, 198). No mentions of unrelated contemporary technologies were found.

49. Authors: added Edqvist’s title, and the annex’s statement that the commission’s work related to Sweden’s accession derogation (p. 195); completed Pedersen’s quoted role and added his 2000 post (p. 198).
50. Absent voices: “scientific case for growth promoters” narrowed to the case for their safety (benefits are acknowledged by the authors and by LBS 1977); SCAN is summarised in two sentences, not one.
51. 9.1.1: “claimed effects” changed to the chapter’s “observed effects”.
52. 9.2: restored “growing concern over” in the Swann trigger quote; the tylosin recommendation is marked as the chapter’s one example of several drug-specific recommendations.
53. 9.3.1 Rebuttal: named the three bans in the quote, and replaced the claim that the p. 96 studies do not bear on the dose mechanism. DVL 1995’s growth-promoter selection finding directly contradicts Walton’s practical claim; the rebuttal is uncited where made, not unsupported (propagated to Mechanism 3(c), Insight 2, Limitations and the digest).
54. 9.3.2: the 1977 group’s proposal was “legislative changes, especially in the requirements for approval” (not “stricter” requirements); added its acknowledgement of disease-prevention benefits; labelled “proximate trigger” as my reading.
55. 9.3.2: the tonnage figure is now flagged as sourced to SOU 1997 and as apparently total animal use.
56. 9.3.2: “risk-maker or risk-taker” restated as the chapter’s “who would bear the costs” (not “should”).
57. 9.3.3: DVL’s selection finding now quoted with “as a growth promoter”.
58. 9.3.7: added WHO 1997’s premise (“Increased concerns regarding risks to public health”).
59. Table 9.1: the same-effect pairings are now within the correct rows.
60. Table 9.2: added discrepancies for 1977 (risk versus impact) and 1985 (date; resistance “in animals”; omitted grounds), and the omission of SCAN 1996 and Copenhagen 1998.
61. References: flagged SOU 1997 “p. 132” as probably the series number SOU 1997:132 (to verify).
62. Authors’ lesson 2: restored “mainly”.
63. Mechanism 2: the directive made the absence of treatment or prevention effects a condition of authorisation; it did not “assume” it.
64. Mechanism 4: “SOU showed” changed to “SOU set out” (it outlined the research needed; it proved nothing).
65. Insight 9: removed the unsourced “small, reputation-sensitive” characterisation of the Nordic sectors.
66. Editors’ amplification: corrected the p. 173 attribution (“as the Swann Report did” refers only to stating whether action precedes research); added the editors’ p. 173 claim that Swann anticipated environmental and welfare problems, which goes beyond Ch. 9.
67. Digest: “Official reassurance” changed to “Reassurance” (Walton was not an official); the Walton rebuttal, DVL basis, tonnage source, “mainly”, Table 9.2 discrepancies and the editors’ p. 173 amplification are aligned with the notes.