LL2-16 digest — Ch16 Seed-dressing systemic insecticides and honeybees#

Late lessons from early warnings II (EEA 2013), pp. 369–406 (PDF 371–408). Authors: Laura Maxim and Jeroen van der Sluijs. The chapter is followed by a Bayer CropScience panel by Richard Schmuck (pp. 401–402) and the authors’ reply (pp. 403–406).

Core story#

In 1994 Gaucho (imidacloprid) was first used as a sunflower seed-dressing in France. That summer, beekeepers saw foragers vanish and disoriented bees, followed by winter losses. According to beekeepers’ reports, sunflower honey yields in affected apiaries fell 40–70% in the following years (pp. 372–373).

Bayer had marketed Gaucho as confined to the soil and even “reducing aerial pollution” (p. 371). Its early studies found no risk and could not detect residues below 10 ppb (p. 373). When public research began, the Ministry’s food directorate (DGAL) told researchers not to go below that 10 ppb limit, which matched Bayer’s method. Bayer also sat on the protocol committee (p. 373).

Public scientists then found residues of a few ppb in pollen and nectar, and sublethal effects at 3–12.5 ppb (pp. 374–376). The official Commission for Toxic Products (CTP) issued ambiguous advice from 1997 to 2002 (p. 379).

In January 1999 the Minister suspended sunflower use under the precautionary principle, on “suspicions … without formally proving” responsibility (p. 374). Maize, which carried far larger commercial stakes, followed only in 2004 (p. 382). That came after repeated ministerial refusals, two State Council rulings that the Ministry had failed its legal duty to assess harm (p. 380), and a 2003 finding by the expert committee (CST) of serious risk to nurse bees (pp. 380–381).

The chapter frames the case as a study of how “substantive”, “procedural” and “social” quality of knowledge shaped a controversy (pp. 371–372). It is not primarily an ecotoxicology review.

Key evidence#

• Methods: acute LD50 and hazard-quotient tools designed for sprays were used for systemic seed-dressings. Those tools miss chronic, sublethal and colony-level effects (pp. 375, 377).
• Persistence: Bayer’s own dossier gave soil half-lives of 188–249 days, beyond the Directive 91/414 trigger (p. 376). Residues appeared in untreated succeeding crops (p. 387).
• How the regulator framed the question: the CTP concluded it could not show that Gaucho maize seed-dressing was “solely responsible, at national level, for all” losses. In the authors’ words, it was answering “a question that had never been asked” (p. 379).
• Regulatory capacity: according to the head of the pesticide bureau (2003), three officials handled 20,000 applications a year, with risk assessment “performed jointly with the industry” (pp. 382–383).
• Pressure on researchers: anonymous testimonies describe legal threats, approaches to managers and cancelled programmes (p. 378). Bayer sued three beekeeper leaders for “discrediting” Gaucho and lost all three cases (p. 380).
• New exposure routes after 2004: sowing dust, guttation drops, widespread residues in pollen, and synergy with the Nosema pathogen (pp. 387, 393, 404–405).
• Other countries: the EU approved imidacloprid in 2008 on a German assessment that, the authors say, omitted French exposure studies (p. 388). Italy’s moratorium (in place by 2009; start date not given) was followed by a drop in reported poisonings from 185 to 2–3, according to its APENET research programme (pp. 393, 404).
• Costs: beekeepers bore them. Beekeeper numbers fell from 84,800 to 68,800, though the chapter calls the insecticides’ share of that decline “unclear”. Research and legal fees came out of development funds, and there was no compensation (p. 383). National average yields show no sustained drop after the bans, apart from a 2005 maize dip attributed to drought (Fig. 16.3, p. 385). Such averages cannot detect localised pest damage.

Authors’ lessons (pp. 388–392)#

Mistrust, fed by poor access to information, drove the conflict. Monitoring needs the trust of practitioners and must draw on lay knowledge. The eight lessons: 1. Check whether existing assessment methods fit a new technology. 2. Develop tests for chronic, sublethal and colony-level effects. 3. Resource regulators properly. 4. Secure expert independence and transparency. This covers publication guarantees, open dossiers, a cost-free database of studies, and public conflict-of-interest declarations. 5. Communicate reliably, completely and responsively. 6. Use explicit quality criteria for selecting studies. Good Laboratory Practice (GLP) guarantees traceability, not quality. 7. Prioritise specific causes rather than letting multicausality excuse inaction. 8. Protect early-warning scientists.

The authors conclude that imidacloprid is “fit for the precautionary principle” given highly variable effects that depend on the environment (p. 392).

Main mechanisms#

• The harm (vanishing bees) was invisible to tests built for acute mortality from sprays.
• Detection limits and question framing produced “no evidence”.
• The fight over field versus laboratory studies was in effect a fight over the burden of proof, and a court settled it (p. 378).
• Expert composition and literature selection shaped conclusions (pp. 379, 386–388).
• On the authors’ account, commercial stakes shaped the order of action (p. 382).
• Prophylactic, bundled deployment led to lock-in, and after the bans other neonicotinoid seed-dressings were put forward (pp. 383–385, 387).
• Costs fell on beekeepers (p. 383), and courts acted as correctives (pp. 380–381).

Transferable insights (with strength)#

1. Legacy assessment tools applied to a new exposure mode create blind spots (pp. 377, 389). Strong.
2. Whoever sets measurement thresholds shapes what counts as “no evidence” (pp. 373–374). Strong for the sequence of events; intent is not shown.
3. Framing the question as “sole cause, everywhere” guarantees inconclusive results (p. 379). Strong.
4. Heterogeneous, unstandardised evidence lets each side assemble its own science (pp. 376, 392). Strong.
5. With no monitoring baseline, neither harm nor the effect of an intervention can be established (p. 386). Strong.
6. Persistent, mobile agents reveal new exposure routes only after wide deployment (pp. 376, 387, 404). Strong.
7. Under-resourced assessment co-managed with applicants yields ambiguous, delay-favouring advice (pp. 379, 382). Moderate.
8. Expert composition and literature selection steer conclusions (pp. 379, 386–389). Moderate.
9. Multicausality used as a reason for inaction; the counter is to prioritise tractable causes (pp. 391, 403). Moderate.
10. Practitioners detect harm first, and excluding them degrades monitoring (pp. 372–373, 388). Moderate.
11. Pressure on independent researchers chills early-warning science (pp. 378, 392). Moderate that it happened; suggestive that it is systemic.
12. Prophylactic, bundled deployment creates lock-in (pp. 383–385). Moderate.
13. The costs of contesting evidence fall on the harmed party (p. 383). Moderate.
14. Action comes first where commercial stakes are lowest (p. 382). Suggestive.
15. Restricting single products leads to substitution by near-equivalents (pp. 387, 393). Suggestive.

Main caveats#

• Explicitly selective: the authors choose data “to reflect the French debate”, and post-2004 sources only from France or banning countries (pp. 371–372).
• Partisan and anonymous sources: heavy reliance on beekeeper organisations, press, a documentary film, personal communications, a second-hand NGO critique and an author-compiled index (pp. 372, 378, 384, 388).
• Confounded attribution: fipronil (Régent TS) was authorised on the same crops from December 1995 and banned together with maize Gaucho in 2004. Varroa, markets and weather also varied. Honey yields kept falling after the 1999 sunflower ban (Fig. 16.1).
• Scale double standard: the authors fault regulators for national-scale framing of bee losses but use national yields to counter farmers’ claims of ban-related losses (p. 385).
• Suchail result unexamined: the Suchail (2001) chronic toxicity finding, one to two orders of magnitude below other values, is not examined for replication.
• One-sided bibliography critique: CST and AFSSA 2009 barely overlapped in either direction, but the authors use this only against AFSSA (p. 386).
• Bayer’s evidence partly unanswered: the printed reply does not engage with the monitoring studies Bayer cites as finding no correlation (pp. 401, 404–405). The longer online version was not checked.
• Internal inconsistencies: ban dates, figure labels and Italian figures (pp. 374/381, 382/399, 384–385, 393/404).
• Bayer’s panel author co-authored the disputed studies.
• Fair to the authors: the critique of the assessment method was later mainstreamed (outside knowledge, to verify). The record of regulatory ambiguity rests on primary official texts. They credit Bayer’s 2002 exposure statement as “a major step forward” (p. 381). Bayer’s aphid argument is weak.