Cavieres, MF, J Jaeger and W Porter. 2002. Developmental Toxicity of a Commercial Herbicide Mixture in Mice: I. Effects on Embryo Implantation and Litter Size. Environmental Health Perspectives 110: 1081-1085

press coverage: LA Times

Cavieres et al. report that a commonly-used commercial herbicide mixture of 2,4-D reduces the litter size of mice exposed to low, environmentally-relevant doses. They found the largest reductions in litter size at the lowest dosage level used, which corresponded to EPA's "reference dose," the concentration calculated on the basis of experiments to be sufficiently low to avoid adverse health effects. This was one-seventh of the "maximum contaminant level" allowed in drinking water by EPA standards.

Developed using traditional toxicological approaches, current EPA standards clearly are inadequate to protect health and require revision. Interviewed about the adequacy of current standards, the research team's senior scientist Dr. Warren Porter (University of Wisconsin, Madison), observed "Federal testing guidelines don't cover real world ingredients, exposure levels, and timing."

What did they do? Cavieres et al. put different amounts of a commercial mixture of 2,4-D, mecoprop and dicamba, plus inert ingredients, in the drinking water of pregnant mice. Impregnated mice were all the same age.

They exposed the mice on two different schedules, (1) the first day of pregnancy (called GD0, or gestational day 0) through to GD15, or (2) GD6 to GD15. The first schedule began the exposure prior to implantation of the embryos and carried it through to the end of organogenesis, the period during which major organs within the embryo take form. The second schedule restricted exposure to organogenesis.

Four exposure levels of the mixture were used and compared statistically to the control group, which received water only. The levels were created by diluting the mixture with varying amounts of water.

After birth, they checked each litter for size, weight of pups, and size of pups. After weaning the mothers were sacrificed and examined by dissection to allow determination of the number of embryo implantation sites. A standard number of pups were left with the mother through weaning and then examined at the age of six weeks for immunological, endocrinological and behavioral attributes. (These observations will be reported in a subsequent paper.)

Because these experiments took place over a period of two years, Cavieres et al. used statistical methods to examine seasonal variation in responsivity to the herbicide mixture.

What did they find? The most important finding is that very low levels of the herbicide mixture caused a large reduction in litter size, indeed larger than the effect of higher doses.

This figure is from Cavieres et al. Graph A (upper left) shows the average litter size of different treatment groups compared to the controls. The asterisk indicates statistical significance compared to controls (p < 0.05). The lowest dose caused the largest effect, in a classic non-monotonic dose-response relationship.

Graphs B through F show the distribution of litter sizes for each treatment group, including controls (B). Note that all treatment groups had at least some litter sizes smaller than smallest ever seen in the controls.

Cavieres et al. also found that the details of the herbicide's effect varied from season to season. While overall, doses caused a decrease in litter size of approximately 20%, in some seasons the largest effect occurred in the high dose groups while in other seasons it occurred in the very low and intermediate dose groups.

Cavieres et al. took a closer, histological look at a subset of the females to determine the numbers of embryonic implantations that had taken place following fertilization. They then compared this with the litter sizes of those same females to calculate the numbers of fetal resorptions that occurred. Comparing controls with treatment groups, litter size and the number of implantation sites did differ significantly in this subset, but resorption rate did not. Thus it appears that the mechanism by which litter size reduction takes place occurs prior to and during implantation, not via fetal resorption.

What does it mean? These results are profoundly challenging to conventional regulatory toxicology.

Traditional toxicological testing of the three herbicides in this commercially available herbicide mixture had characterized them as safe at the lowest levels used in these experiments. Those standards were based upon experiments carried out with pure forms of the herbicides one-by-one. Here, Cavieres et al. have carried out experiments that come much closer to the real world situation in which people and animals will encounter these herbicides. They find significant adverse effects at the lowest levels they tested. Clearly, something is wrong with the conventional approach.

Given the way Cavieres et al. conducted their research, it is impossible to determine the precise mechanism or even the agent of toxicity. Was it the 2,4-D in combination with the other herbicide compounds, or perhaps with one of the "inert ingredients" present in the commercial herbicide mixture? Or was dicamba the principal agent? They can't say. Nonetheless, somehow the mixtures they used caused effects at levels previously deemed safe for individual components.

To get at the precise contributions of different components will take considerable effort and resources, and to do this for all possible combinations of herbicides and inert ingredients in commercial use would be physically and financially impossible.

Yet conventional regulatory toxicology begins with individual components, because of the precision with which experiments can be conducted, and never gets to real world mixtures. If Cavieres et al.'s results are at all indicative of what happens with other mixtures, then current regulatory standards are unlikely to be sufficiently protective of public health. Indeed, previous epidemiological research on related herbicides indicate adverse effects in areas of high use.

The results presented by Cavieres et al. argue strongly that testing of pesticides for regulation should focus on the mixtures of compounds that people and animals encounter in the real world. The conventional approach sacrifices accuracy for reductionist precision. An more constructive regulatory approach would be to prioritize testing based on two considerations: focus first on the commonest commercial mixtures and on mixtures detected through direct measurements of body burden.

These results also offer yet another example of non-monotonic dose response curves, violating the simplistic interpretation of that oft-heard basic tenet of toxicology, that the dose makes the poison. Yes, the herbicide's effect did vary with dose. But the usual way this assumption is used is to imply that high doses have greater effects than low doses. In this case, that is clearly not true.