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

Late lessons from early warnings (EEA, 2001), pp. 93–100. Authors: Lars-Erik Edqvist (Director General, Swedish National Veterinary Institute; chaired the Swedish Commission on Antimicrobial Feed Additives, 1995–97, p. 195) and Knud Børge Pedersen (Director, Danish Veterinary Laboratory, p. 198). No panels.

Core story#

• Discovery and adoption. In the late 1940s, residual tetracycline in feed waste was found to speed chicken growth (Stokstad and Jukes 1949). Low-dose antibiotic feeding was “readily adopted” and became integral to industrial husbandry. Growth promoters were sold over the counter, while therapeutic antimicrobials were in most cases prescription-only (p. 93).
• Resistance was already known. It was well acknowledged by the early 1950s, and horizontal gene transfer was shown in the 1960s (pp. 93–94).
• Swann warns. After growing concern over multi-drug-resistant Salmonella, the UK Swann Committee (1969) judged the evidence “a sufficiently sound basis for action”. It proposed restricting non-prescription feed antibiotics to those of economic value, with “little or no” therapeutic use, that would not impair therapy through resistance. It also proposed keeping tylosin prescription-only and creating one cross-sectoral committee for all antimicrobial use (p. 94).
• Dilution. Most recommendations were adopted, then “gradually diluted”:
• the cross-sectoral committee “not fully implemented”, and no resistance monitoring;
• the EU approved tylosin and spiramycin as growth promoters in 1975;
• avoparcin use expanded as its medical equivalent vancomycin entered hospitals (p. 94).
• Reassurance. It rested on models later overturned. In 1977 an official Swedish Board of Agriculture working group accepted a resistance risk but held that resistance from gram-positive bacteria “is not transferred”, so the impact was “negligible” (p. 95). In 1988 the scientist Walton (no affiliation given) argued that sub-inhibitory doses cannot select resistance, so Swann was “in error”. The chapter’s rebuttal is uncited where it is made (pp. 94–95), though DVL 1995’s finding that growth-promoter use of avoparcin selects VRE (p. 96) contradicts Walton.
• Sweden first. Swedish farmers, worried about “consumer confidence”, asked for a ban in 1984. The government cited resistance risks and long-term “uncertainty”, and Sweden banned growth promoters from 1986. Total (apparently animal) antimicrobial use fell from ~50 to ~20 tonnes by 1996, a figure sourced to SOU 1997 (p. 95).
• Denmark and the EU. When VRE in animals was reported in 1994–95, Danish farmers and the feed industry stopped avoparcin voluntarily “to reduce the spread of antimicrobial resistance”, and the government banned it in May 1995 (p. 96). The EU followed in April 1997, although its scientific committee (SCAN) said further evidence was needed to establish a risk, while accepting that “serious questions” had been raised. Denmark banned virginiamycin in 1998 to protect the new human drug Synercid. The EU banned four more growth promoters from July 1999 as “a precautionary measure”, and the virginiamycin manufacturer sued (p. 96). WHO, the EU Scientific Steering Committee (SSC) and a Copenhagen conference urged replacement or phase-out (p. 97).

Key evidence#

• Scientific basis for Danish action: cross-resistance, transferability, selection of VRE by avoparcin used as a growth promoter, and VRE reaching humans via food (DVL 1995, p. 96).
• Three cross-resistance pairs: avoparcin/vancomycin, virginiamycin/Synercid, avilamycin/everninomycin (pp. 94–97).
• The SOU 1997 assessment:
• risk “far from negligible”;
• establishing the risk would need research along a 17-step causal chain, taking at least 5–10 years, done for each resistance gene and each substance, with later updates;
• inaction “is not a neutral position” (pp. 95–96).
• No quantified human harm. WHO (1997) said the magnitude “is not known” (p. 97).

Authors’ lessons (p. 98)#

• Swann rested on “a low level of scientific proof” but “competent microbiological assessment”, and was “accurate” and “far-sighted”.
• The dilution relied “mainly” on “what was precisely known rather than … what was not known”. Science needs “more humility and less hubris”.
• Committees assessing confidential industry data need broad, independent, multidisciplinary membership, especially human medicine.
• Risk assessments should be wider, including alternatives such as better husbandry.
• Under uncertainty, caution is preferable. This repeats SOU 1997 almost verbatim; the authors add that precaution is needed especially where waiting for “ultimate proof” risks deaths.
• Farmers’ voluntary action and Swann were “vindicated by history”.

Main mechanisms#

• Lock-in and masking: the additive enabled intensive systems whose problems it then hid (p. 93; Table 9.1).
• Purpose-based classification: feed additive versus medicine, assessed by an animal-nutrition committee (pp. 93, 95–96).
• Shifting reassurance: confinement to the clone, then to gram-negative bacteria, then a low-dose argument, each overturned (pp. 94–95, 98). The “retreat” reading is the notes’, not the chapter’s.
• Slow-to-meet proof standard and calls for research that in effect let use continue (pp. 94–96).
• Monitoring never set up; industry-supported studies outnumbering independent ones (pp. 94, 97).
• Split interests: Swedish farmers broke ranks over consumer trust, and Danish farmers and feed makers halted avoparcin over resistance, while the drug supplier litigated (pp. 95–96).
• Frontrunner states using EU derogation and safeguard clauses (pp. 95–96).
• Costs on outsiders: they fall on patients, health systems and future drugs (“risk-maker or the risk-taker?”, p. 96; Table 9.1).

Transferable insights (technology-neutral)#

1. Regulating by stated purpose rather than function can exempt an agent from controls applied to the same agent elsewhere (pp. 93, 95). Moderate.
2. Reassurances can rest on mechanistic models that treat the edge of knowledge as a boundary of risk, shifting ground as knowledge moves (pp. 94–95, 98). Moderate (two documented reassurances, 1977 and 1988; the dose rebuttal is uncited where made, though supported by DVL 1995, p. 96).
3. When proof needs a long causal chain for every variant, “prove harm first” in effect means years of continued use, reopened with each new variant; inaction is a decision (pp. 95–96). Moderate.
4. Monitoring that is never set up produces an absence of evidence (pp. 94, 97). Moderate for the gap; Ch. 9 does not show the gap being used as reassurance.
5. Research follows interest; side effects stay under-studied until a sentinel event (p. 97). Suggestive.
6. Warnings can erode through exceptions and incomplete machinery rather than repeal (p. 94). Moderate/suggestive (one case).
7. Downstream users may act before regulators (pp. 95–96). Moderate. The public-trust motive is documented only for Sweden; the Danish halt cited resistance.
8. Masking technologies deepen dependence and suppress alternatives (Table 9.1). Asserted.
9. Use of a shared, depletable resource in one domain can pre-emptively erode future options elsewhere (pp. 94, 96–97). Moderate.
10. The long lag comes before specific pathway evidence. Once a sentinel finding appears, action can take months (p. 96). Strong (timeline). Sweden acted earlier without such evidence, and the EU acted despite SCAN’s doubts.
11. Unknown magnitude is not unknown direction; action can be coherent without size estimates (pp. 95–97). Moderate.
12. Committee composition shapes which harms count (p. 98). Asserted.

Main caveats#

• Insider authors. The key sources (SOU 1997, DVL 1995) are the authors’ own commission and institution. This is disclosed only in the author annex, not in the chapter. The concluding principle restates SOU text.
• Vindication is mechanistic and regulatory, not outcome-based. There are no figures on illness, deaths or costs, and the “substantial evidence” claim carries no citation (p. 97).
• Counter-arguments are thin. The scientific case that growth-promoter use was safe appears only through Walton (whose affiliation is not given), and its rebuttal is uncited where made. No industry study is described. SCAN’s reasoning is barely described. Post-ban animal health, productivity and therapeutic use go unexamined. Scientific advice before the 1998 ban is omitted.
• Internal wrinkles. The chapter dates cross-genera transfer to both the 1960s (p. 94) and the 1990s (p. 98). The Swedish Bill’s “prevention” sits uneasily with the “therapy only” summary (p. 95).
• Table 9.2 (EEA-compiled) diverges from the text. It sharpens WHO’s position, garbles 1975, calls the 1977 risk (not impact) negligible, dates the Swedish ban to 1985 and narrows its grounds to resistance “in animals”, adds “health” to the Swedish farmers’ motive, and omits the avoparcin bans, the Danish virginiamycin ban and SCAN’s 1996 opinion.
• The editors’ synthesis (pp. 170–182) attributes more to this case than Ch. 9 shows, including on innovation, public misgivings, non-compliance, and Swann’s anticipation of environmental and welfare problems (p. 173).
• Hindsight checks: the court outcome, EU and US follow-through, Danish post-ban data and critiques, and low-dose selection research.