The Study Nobody Has Run

Under the US EPA's residue chemistry test guidelines, if a pesticide residue on tobacco is found at or above 0.1 parts per million, the registrant must conduct a pyrolysis study: burn the material, identify the combustion products, and submit them for evaluation of human health significance.

The reasoning is straightforward. Tobacco is burned and inhaled. Therefore the combustion products, not merely the residue, are the relevant exposure.

The requirement has existed since at least 1996. There is no equivalent for any other inhaled plant material, anywhere in the world.

The regulatory system already contains the correct scientific instinct — that when a residue will be burned and breathed, somebody must find out what burning produces.

It has been applied to exactly one crop. And the threshold that triggers it has since been borrowed and repurposed as a standalone limit for other inhaled material, stripped of the requirement that gave it meaning. California's own memorandum documents this, describing 0.1 micrograms per gram as the level "that triggers the US EPA to require pyrolysis testing" — and then adopting it as a default action level in a context where no such testing occurs.

A number that meant *go and investigate* now means *you are compliant*.

It is not exotic. The protocol has existed for decades and has been executed repeatedly for other compounds: apply a known quantity of the substance to plant material; pyrolyse it in a standardised tube furnace across a range of temperatures spanning realistic combustion; trap the effluent; identify products by gas chromatography–mass spectrometry against authentic standards; quantify against calibration; establish mass balance.

Roughly a week of instrument time.

That protocol was used in 1979 to characterise a herbicide's combustion products on plant material, with confirmation in genuinely contaminated field samples. It was used in 2011 on an insecticide, with full quantification. It was used in 2012 on a growth regulator on tobacco — that study found only 1.4 to 3.7 per cent reached mainstream smoke and that most was destroyed. It has been applied to fungicides, to dithiocarbamates, to industrial materials of every description.

It has never been applied to paclobutrazol. Or daminozide. Or chlormequat. There is no thermal analysis, no pyrolysis chromatography, no tube-furnace study and no combustion-product identification for any of them, at any temperature, on any substrate.

In the absence of that work, three kinds of claim circulate.

The first is measurement. A 2013 study found that 49 to 67 per cent of a growth regulator applied to plant material was recovered *intact* in mainstream smoke. That is real data — though the authors worked for a commercial testing laboratory, the material was artificially spiked at levels well above realistic contamination, and a 2022 tobacco study warns explicitly that spiked substrates overstate transfer. Treat it as an upper bound. Note too that the paper's own conclusion is that these compounds are "highly resistant to pyrolysis" — the finding is that they survive, not that they transform.

The second is inference from analogous chemistry: the triazole combustion products discussed elsewhere in this cluster, or the well-characterised behaviour of quaternary ammonium compounds under heat. Legitimate, if labelled as inference.

The third is assertion. Claims about specific carcinogenic combustion products of specific growth regulators circulate widely and, when traced, resolve to safety-data-sheet boilerplate generated from elemental composition, or to a real finding about one compound transferred onto a chemically different one. Some of it traces to summarisation tools that invented citations outright.

The honest summary is uncomfortable for everyone holding a strong position.

The alarmed view overstates what has been measured. The reassuring view — that residues sit within limits — rests on limits that were never health-based. And the experiment that would resolve it costs about a week of instrument time and has gone unfunded for four decades.

Until somebody runs it, the only input specification that requires no assumptions about combustion chemistry is one that contains nothing to combust.