In the present review we focus on the potential for neurotoxicity after developmental exposure to pyrethroid insecticides. We also consider the current state and quality of scientific data that could be used to support risk decisions related to pyrethroid developmental and age-dependent neurotoxicity. Specifically, in this review we a) provide a brief overview of the toxicity of this class of compounds; b) review pyrethroid effects on voltage-sensitive sodium channels (VSSCs), a primary mode of action of pyrethroids; c) discuss the developmental profiles of VSSCs; d) provide examples of the results of perturbation of VSSCs during development by other insults; e) discuss the evidence for age-related sensitivity to this class of compounds; f ) summarize and critique studies of pyrethroid neurotoxicity after developmental exposure; and g) make recommendations regarding future research needs related to the developmental neurotoxicity of pyrethroids.

In addition to being important to scientists interested in characterizing the neurotoxicity of these compounds, this information will be useful when considering the scientific data needed to inform risk decisions related to pyrethroid insecticides. Under the Food Quality Protection Act (FQPA; 1996), the U.S. Environmental Protection Agency (EPA) is required to include a default 10× safety factor (uncertainty factor) in risk decisions to protect against potentially greater sensitivity of developing individuals to toxic insult. This factor can be adjusted only if compelling scientific data exist regarding age-related differences in sensitivity. Furthermore, developing individuals must be considered under FQPA requirements for cumulative risk assessments (classes of compounds with the same mode of action). The quality of the scientific data used to support these and other risk decisions is an important component of scientifically based risk assessment. In addition, information regarding mode of action improves the scientific basis for risk decisions (Brock et al. 2003; Mileson et al. 1998; Sonich-Mullin et al. 2001), including those related to developmental neurotoxicity (Costa 1998; Tilson 2000a, 2000b).

The U.S. EPA has recently released the revised cumulative risk assessment of the organophosphate class of insecticides (U.S. EPA 2002) and has requested that registrants of these insecticides submit developmental neurotoxicity studies to the agency. In the near future, the U.S. EPA must consider developmental and cumulative risk for other classes of insecticides, including pyrethroids. Thus, in this review we focus on issues of mode of action and age-dependent and developmental neurotoxicity as related to risk decisions under the FQPA.

Knockout and mutant mouse models of sodium channel α subunits demonstrate varying degrees of adverse outcomes associated with loss or alteration of specific channel subunits. When mRNA for the Na_v1.2 subunit was reduced by approximately 85%, mice exhibited reduced levels of electrical excitability, had high levels of apoptotic neurons in the brainstem and cortex, and died from severe hypoxia within 1–2 days of birth (Planells-Cases et al. 2000). In contrast, mutation of the gene encoding the Na_v1.6 subunit resulted in development of hindlimb paralysis, skeletal muscle atrophy by postnatal day (PND)10, and death by PND20 (Porter et al. 1996). Atrophy was specific to muscle innervated by spinal and not oculomotor neurons (Porter et al. 1996). Finally, Na_v1.8 knockout mice survived to adulthood and exhibited normal behavior, although sensation of some types of noxious stimuli was lost or diminished (Akopian et al. 1999; Laird et al. 2002).

In humans, perturbation of nervous system development has been associated with altered VSSC structure or function. Recent advances in molecular genetics have identified in genes coding for VSSC subunits a number of mutations that result in neuronal hyperexcitability due to subtle changes in channel gating and inactivation (see Meisler et al. 2001, their Table 3). These mutations have been linked to various forms of epilepsy in humans, providing evidence that changes in VSSC function can give rise to clinically definable disease (Claes et al. 2001; Escayg et al. 2001; Meisler et al. 2002; Noebels 2002; Wallace et al. 2001). Mouse models expressing these mutant ion channels have been constructed, facilitating the study of these diseases (Kearney et al. 2001; Meisler et al. 2001). It is noteworthy that pyrethroids, like these mutations, alter VSSC activation, inactivation, and neuronal excitability. The mechanisms and magnitude of mutational versus pyrethroid effects are different, as would be the duration of effect (dependent on exposure for pyrethroids vs. permanent for mutations). Because of these differences, results from mutation and knockout models may not be predictive of developmental exposure to pyrethroids. Notably, potential interactions between pyrethroids and these mutations to VSSCs have not yet been examined.

Allethrin (in the form of allethrin, d-allethrin, bioallethrin, and S-bioallethrin) and permethrin are the only type I pyrethroids for which peer-reviewed studies of potential developmental neurotoxicity have been conducted. Of the type II compounds, results of developmental studies have been published for deltamethrin, cypermethrin, fenvalerate, and cyhalothrin, and data regarding the developmental neurotoxicity of cyfluthrin (Jekat et al., unpublished data) have been submitted to the U.S. EPA. Thus, no developmental neurotoxicity studies exist for many pyrethroids.

Rodents were the sole animal models used in these studies: 13 studies used rats and 9 studies used mice. No studies were conducted specifically to examine species differences, nor could any clear species-dependent effects be discerned. The choice of rats or mice seemed to be based on a) previous use of that species in the laboratory or b) whether or not the study was designed to replicate (in whole or part) results published previously by other investigators. A systematic comparison of factors that underlie potential species differences in neurotoxic responses could provide useful information regarding the extrapolation of data from animals to humans. For example, Na_v1.3 expression in rodents appears to be primarily embryonic, yet in humans considerable expression in adults has been reported (Whitaker et al. 2000, 2001). How this and other species differences influence neurotoxic responses has not been investigated.

Several studies reported persistent changes in behavior and/or neurochemistry in animals examined long after exposure had stopped. Eriksson’s group (Ahlbom et al. 1994; Eriksson and Fredriksson 1991; Eriksson et al. 1993; Eriksson and Nordberg 1990) has reported that mice exposed to pyrethroids during PND10–16 exhibit increased motor activity and lack of habituation, as well as changes in density of muscarinic acetylcholine receptor (mAChR) binding for as long as 5 months (Talts et al. 1998) after cessation of exposure. Given the short half-lives for pyrethroids (Anadón et al. 1991, 1996; for review, see Kaneko and Miyamoto 2001), these effects are likely due to exposure during development and not residual tissue concentrations of pyrethroids. Studies conducted by Eriksson and co-workers used bioallethrin and deltamethrin, which contain only two and predominantly one stereoisomer, respectively. Thus, effects can be ascribed to the compound that has insecticidal activity (vs. studies conducted with formulated products). In addition, dose–response relationships have been demonstrated for bioallethrin (Ahlbom et al. 1994), and the replication of effects, both behavioral and biochemical, within this laboratory has been consistent over several studies. Others have also reported persistent changes in behavior and/or biochemistry, including learning (Moniz et al. 1990), motor activity (deltamethrin only; Husain et al. 1992), sexual behavior (Lazarini et al. 2001), mAChR binding (Aziz et al. 2001; Malaviya et al. 1993), and blood–brain barrier permeability (Gupta et al. 1999a).

There were several studies that examined both motor activity and mAChR expression after developmental exposure to pyrethroids. A summary of effects on these end points, independent of dose, exposure period, and other parameters, is provided in Table 5. In all of these studies, quinuclidinyl benzilate (QNB) binding was used to measure mAChR expression. QNB is a nonspecific antagonist for this receptor (Watling et al. 1995) and does not discriminate between mAChR subtypes (M1–M5). Measurement of QNB binding may in fact be one of the more comparable end points across these numerous studies. In addition, many but not all of these studies examined mAChR expression at PND17 and/or 4 months of age.

Comparison of pyrethroid effects on QNB binding across studies does not reveal clear trends in reported effects between laboratories. In preweanling animals, across all compounds and treatment protocols, QNB binding was reported to increase in six studies, decrease in two studies, and not change in four studies (Table 5). In cortical tissue, the data for PND17 are more consistent in that five of eight studies reported increases in mAChR expression. If only the various forms of allethrin are considered, four studies reported increases and two reported no change in QNB binding when measured on PND17. Persistent alterations in mAChR in adulthood after developmental exposure are less clear, with three studies reporting increases, three reporting decreases, and five reporting no change in QNB binding. Considering only allethrin forms again, QNB binding increased or decreased in two studies each and was unchanged in three studies.

Differences in a number of important variables may underlie some of the inconsistencies in QNB binding data. One difference is exposure route. Two studies used inhalation exposure (Ivens et al., unpublished data; Jekat et al., unpublished data), whereas exposure in the remainder of the studies was via oral gavage (Table 5). A comparison of effects in Tables 2–5 suggests that this is not a tenable explanation for these inconsistencies because results do not correlate to route. Another variable that differed between laboratories was the formulation of allethrin used. Allethrin, like all pyrethroids, exists as several different stereoisomers (Figure 2), and the insecticidal and toxic effects of pyrethroids are highly stereospecific. These studies employed allethrin formulations with differing contents of allethrin stereoisomers; two groups used d-allethrin (Ivens et al., unpublished data; Tsuji et al. 2002), one used bioallethrin (Eriksson group: Ahlbom et al. 1994; Ericksson and Fredriksson 1991; Eriksson and Nordberg 1990; Talts et al. 1998), and two used S-bioallethrin (Muhammad and Ray, unpublished data; Pauluhn and Schmuck 2003). Again, data in Table 5 suggest that this is not a tenable explanation because d-allethrin and bioallethrin result in either increases or no effects on mAChR binding. An additional variable in these data sets is the specific methods used in the competitive binding experiments. Competition experiments with carbachol were used in several studies to distinguish between high- and low-affinity QNB binding sites (Ahlbom et al. 1994; Eriksson and Fredriksson 1991; Eriksson and Nordberg 1990; Ivens et al., unpublished data; Jekat et al., unpublished data; Talts et al. 1998). Two studies (Ahlbom et al. 1994; Eriksson and Nordberg 1990) reported that bioallethrin increased the percentage of low-affinity binding sites in PND17 mice, an effect not reported in adult mice, despite changes in the density of muscarinic binding (Eriksson and Fredriksson 1991; Talts et al. 1998). Ivens et al. (unpublished data) did not find changes in the percentages of high- and low-affinity sites, even though they did report changes in the density of QNB binding sites in PND17 animals. In some cases, the relative proportion of high- and low-affinity sites was not investigated even though changes in density were reported (Muhammad and Ray, unpublished data). The ability to distinguish high- and low-affinity sites, and effects thereon, is dependent on the number of points included on the agonist competition curve. Studies conducted by the group at Bayer (Ivens et al., unpublished data; Jekat et al., unpublished data) used seven different concentrations of carbachol, whereas studies conducted by Eriksson’s group (Ahlbom et al. 1994; Eriksson and Fredriksson 1991; Eriksson and Nordberg 1990) used 18 concentrations of carbachol (Eriksson P, personal communication). This information was typically not available to evaluate and may account for some reported differences, because use of too few points may preclude detection of changes in the low-affinity site. Overall, the data across laboratories indicate that changes in QNB binding may not be a robust response to developmental exposure to pyrethroids and that conditions may need to be more carefully controlled in order to observe changes.

A smaller number of studies examined potential alterations in catecholaminergic systems. Both deltamethrin (Lazarini et al. 2001) and bioallethrin (Muhammad and Ray, unpublished data) were reported to increase 3,4-dihydroxyphenylacetic acid (DOPAC) levels in the adult striatum after developmental exposure. However, developmental exposure to a commercial product containing fenvalerate had no effect on monoamine levels in the striatum (Moniz et al. 1999). Malaviya et al. (1993) reported that binding of ³H-spiroperidol to striatal membranes from PND21 rats was decreased and increased, respectively, after gestational and lactational exposure to a commercial product containing fenvalerate, whereas binding was increased after only lactational exposure to a commercial product containing cypermethrin. Thus, similar to the muscarinic cholinergic system, the dopaminergic system may be affected by developmental exposure to pyrethroids, but studies examining this system have reported inconsistent results to date.

Eriksson and co-workers have consistently reported increased motor activity and a lack of habituation after exposure to pyrethroids (Ahlbom et al. 1994; Eriksson et al. 1993; Talts et al. 1998). A comparison of effects of pyrethroids on motor function between laboratories is not as consistent. Muhammad and Ray (unpublished data) observed effects on motor activity in some cohorts but not in others. After inhalation exposure to bioallethrin (Tsuji et al. 2002) or d-allethrin (Ivens et al., unpublished data), no effects on activity or habituation were reported. By contrast, inhalation exposure to cyfluthrin resulted in hyperactivity and decreased habituation in female mice (Jekat et al., unpublished data). Several additional studies also examined other measures of open field or motor activity (Gomes et al. 1991a; Husain et al. 1992, 1994; Lazarini et al. 2001). Reports of effects in these studies were also variable (Table 2). The reasons for the discrepant nature of these findings are unknown.

A small number of studies tested cognitive functions (Table 2). Two studies reported that bioallethrin exposure during PND10–16 (via different routes) had no significant effect on performance in the Morris water maze at 5 (Talts et al. 1998) and 11 (Tsuji et al. 2002) months of age. Other studies reported decreases in avoidance and Y-maze learning (Aziz et al. 2001; Husain et al. 1994; Moniz et al. 1990) or no change in avoidance behavior (Gomes et al. 1991a). A major confounder in the Y-maze and avoidance studies is the use of commercial formulations rather than technical compound.

There are several common weaknesses in the developmental studies that temper the scientific strength of some individual reports, as well as the data set when taken as a whole. A key weakness is problematic statistical analyses. Most behavioral studies [with the exception of Ivens et al. (unpublished data), Jekat et al. (unpublished data), and Tsuji et al. (2002)] used multiple pups from the same litter without correction in the statistical analysis. The sampling of multiple pups from the same litter inflates the sample size and increases the probability of a type I statistical error (Abbey and Howard 1973; Holson and Pearce 1992; Muller et al. 1985; Reily and Meyer 1984). When biochemical end points were examined, statistical analyses often lacked robustness or, in some cases, were absent. In several studies examining receptor binding, results were compared (and significant differences found) using multiple Student’s t-tests. Use of multiple t-tests can easily increase the probability of a type I error (Muller et al. 1985). These study designs should use statistical models that control for multiple comparisons (e.g., analysis of variance with appropriate post hoc test for comparisons of different group means). Meta-analyses or other statistical approaches to examine related data sets from the same and different laboratories could help strengthen conclusions when effect magnitude is small but have not been conducted to date.

An additional limitation common to these reports was a lack of tissue concentration data. None of the studies reported pyrethroid blood or brain concentrations from dams or pups. Such information would have greatly facilitated comparisons between studies and would also be useful to compare target tissue concentrations in the test species with exposure estimates in pregnant women (see Whyatt et al. 2002).

Lack of information about the stereoisomer composition and/or purity of the test compound was a serious confound in some reports. Such information is important to be able to compare studies generated in different laboratories, as discussed above for the different allethrin products. In addition, several studies used formulated products rather than purified compound (Aziz et al. 2001; Gupta et al. 1999a, 1999b; Husain et al. 1992, 1994; Malaviya et al. 1993). Formulated pesticide products typically contain solvents, emulsifying agents, petroleum distillates, and other “inerts” (Farm Chemicals Handbook 1997), many of which are known or suspected to have neurotoxic properties. Although use of formulated products may provide a more real-life exposure situation, lack of information on the content of proprietary formulations hampers comparisons between studies and often precludes attributing effects directly to the pyrethroid.

Several other limitations should also be noted. The number of time points examined in these studies typically was three or fewer, one of which was often a measurement in adult animals. Considerable ontogeny of both behavioral responses as well as biochemical end points is well established. Thus, the tendency of most studies to examine a “snapshot in time” may miss important ontogenic shifts induced by these compounds. Dosing duration and age at exposure are two other important factors. Although a number of studies examined the period of PND10–15, the choice of dosing periods in the present studies was variable, and, to date, there has not been a systematic evaluation of potentially sensitive developmental periods. An additional consideration regarding dosing periods is the differential rates of neurodevelopment in rodents versus humans. Thus, studies such as those conducted by Whyatt et al. (2002) could potentially provide important information about exposure to the developing fetus. In addition, the effects of sex were not always considered in the present studies, with a few exceptions (e.g., Gomes et al. 1991b; Moniz et al. 1999). Also related to this topic is the relative distribution of males and females in a litter. In some cases, culling information was readily available; however, many studies provided no or insufficient information to evaluate this variable.

Biologically based dose–response (BBDR) models (Andersen and Dennison 2001) describe the relationships between different components of the continuum between exposure to and the adverse effects of a chemical (Figure 4). For example, such a model has recently been constructed for the developmental neurotoxicity of perchlorate (Jarabek et al. 2002). Mode-of-action models strengthen science in two important ways. First, the uncertainty regarding animal-to-human extrapolations can be reduced if a toxicant’s mode of action in an animal model is demonstrated to be relevant to humans (Cohen et al. 2004; Meek et al. 2003; Sonich-Mullin et al. 2001). Second, these models often provide insight into research needs by identifying data gaps and research needs. For pyrethroids, much of the future research needs can be described in the context of the type of data that would be useful in constructing a BBDR for this class of compounds, or for individual compounds within this class. A cornerstone of a BBDR model is a physiologically based pharmacokinetic (PBPK) model that describes the relationship between exposure and target tissue dose (Andersen and Dennison 2001). Additional pharmacokinetic information in animal models as well as additional pharmacokinetic and exposure information in humans is needed. For pyrethroids, this will involve defining the relationship between maternal and fetal compartments, and the involvement of oral (including lactation), inhalation, and dermal exposures to the newborn. Current data indicate that some exposure does occur to pregnant mothers, infants, and children, resulting in low internal doses (Berkowitz et al. 2003; Heudorf et al. 2004; Schettgen et al. 2002). However, insufficient information is available to adequately evaluate the range of internal doses of pyrethroids in humans. These data will be valuable in quantitative extrapolations of exposure from animals to humans (Andersen and Dennison 2001). Pharmacokinetic information is available comparing acute high-dose exposures in neonatal versus adult animals (Cantalamessa 1993; Sheets et al. 1994). However, only a limited number of compounds have been examined to date, and no information is available for ages before PND11.