Influence of soil organic matter content on the toxicity of pesticides to soil invertebrates: A review

A better understanding of how soil organic matter (OM) content influences pesticide toxicity to soil invertebrates is needed to improve the ecological relevance of risk assessment approaches. In the current study, soil invertebrate toxicity data (LC50 and EC50 values) were collected from studies determining the toxicity of organic chemicals in soils with varying OM content. Relevant studies were identified by performing a literature search and through the use of toxicity databases. The data were used to address the following questions: (1) Can the relationship between toxicity and soil OM content be quantified? (2) Does soil OM content influence different toxicity endpoints in a similar way? (3) Is the influence of soil OM content on sensitivity to pesticides different between species? The results indicate that toxicity—OM relationships are chemical dependent, differ between endpoints, and are species‐specific. Hence, the grouping of chemicals based solely on their lipophilicity, as well as having only one correction factor for multiple species, may not be an appropriate approach to risk assessment. Integr Environ Assess Manag 2023;19:1457–1472. © 2023 The Authors. Integrated Environmental Assessment and Management published by Wiley Periodicals LLC on behalf of Society of Environmental Toxicology & Chemistry (SETAC).


INTRODUCTION
Soil properties such as organic matter (OM) content, clay content, and pH influence the bioavailability of chemical pollutants to soil organisms, as they determine its partitioning between the soil solid phase and the pore water (Amorim, Römbke, Scheffczyk, Nogueira, et al., 2005;Domene et al., 2012).As such, these properties determine a pollutant's concentration in the pore water, which is considered to be the main route of exposure for soil invertebrates (European Food Safety Agency [EFSA], 2009; Peijnenburg et al., 2012;Van Gestel & Ma, 1990).Consequently, the impact of soil properties on the toxicity of chemicals is a major source of uncertainty in the environmental risk assessment for soil organisms, as the bioavailability (and thus toxicity) of a chemical is dependent on the soil used in the toxicity test.In Europe, first-tier regulatory toxicity tests are mostly performed using standardized Organisation for Economic Co-operation and Development (OECD) artificial soil.For earthworms, these soils consist of 70% quartz sand, 20% kaolin clay, and 10% sphagnum peat, while 75% quartz sand and 5% sphagnum peat are used in studies with Folsomia candida and Hypoaspis aculeifer.Calcium carbonate is added to adjust the pH (1 M KCl) to approximately 6.0 (OECD, 1984).These artificial soils are preferred to natural soils in standardized ecotoxicological studies, as the materials are readily available, allowing different laboratories to create similar soils, resulting in reproducible and comparable results.Additionally, they do not contain any organisms or pollutants, which can influence the outcome of the toxicity test (Brami et al., 2017;Hofman et al., 2009).
The use of OECD artificial soil in first-tier laboratory toxicity tests has been criticized, mostly because the results are difficult to extrapolate to the field due to the lack of similarity to natural soils.Currently, in the European risk assessment of pesticides, a safety factor of 5 is applied to cover various sources of uncertainty when extrapolating results from first-tier laboratory tests to effects in a field situation.For instance, the 10% sphagnum peat in OECD artificial soil does not represent the natural OM type and content generally found in European agricultural soils, which typically contain <5% OM (Chelinho et al., 2014).The main concern is that if natural soils contain less OM, chemicals in these soils will be more bioavailable due to weaker sorption to the soil matrix, and therefore, estimates of toxicity generated using artificial soils could be underestimated.This may especially be the case for lipophilic organic chemicals, as they have the potential to strongly adsorb to soil OM, thus reducing bioavailability.On the other hand, as a result of increased adsorption to the soil, pore water may not be the only route of exposure for soil invertebrates, and soil and litter passing through their gut may become an additional route of exposure (Belfroid et al., 1995), further complicating the extrapolation of toxicity data among soils.To compensate for the differences in bioavailability and thus toxicity, the European Food Safety Authority (EFSA) mandates that when an active substance is lipophilic (log K ow > 2), the endpoints (e.g., LC 50 , EC 50 , EC 10 , NOEC) derived from first-tier chronic earthworm toxicity tests performed in OECD artificial soil should be divided by a factor of 2 to correct for the usually lower soil OM content in agricultural field soils (EFSA, 2017).Moreover, the correction factor is also applied to studies with springtails and predatory mites, although these are often performed in artificial soils with a peat content of 5%, which is similar to the OM content typically found in European agricultural soils.These corrected endpoints are then used in the subsequent risk assessment.
The use of this correction factor of 2 has recently been questioned, since "the assumption of a one-to-one relationship may not be correct, especially as the underlying scientific justification for the use of the correction factor is not well elaborated" (EFSA, 2015).The correction factor of 2 is based on the results of acute 14-day toxicity tests with earthworms focusing on survival and only a selected number of organic chemicals, which makes its applicability to other chemicals and longer-term tests on earthworm reproduction questionable.The applicability to hard-bodied arthropods is equally questionable as these organisms may have different exposure routes to organic chemicals compared to softbodied earthworms (ECHA, 2016).
The goal of this study is to investigate the influence of soil OM content on the sensitivity of soil invertebrates to organic chemicals and in particular pesticides.Through a review and analysis of the existing literature and databases, the following questions are addressed: 1) Can the relationship between pesticide toxicity and soil OM content be quantified?2) Is the influence of soil OM content on pesticide toxicity similar for different toxicity endpoints (i.e., survival and reproduction)? 3) Is the influence of soil OM content on sensitivity to pesticides different between invertebrate species, and can this difference be explained by differences in the route of exposure to chemicals?
To answer these questions, a systematic literature review was conducted in which laboratory-derived soil invertebrate toxicity data were collected from studies determining chemical toxicity in soils with varying OM content.Needs for further research to unravel the relationship between the toxicity of chemicals to soil invertebrates in relation to soil properties are identified, and the applicabilities of OM-toxicity-based correction factors within the regulatory assessment of pesticides are discussed.Finally, the relationship between toxicity and soil OM content is also discussed in the context of the relevance of artificial soils to more natural substrates.

Data collection
Mortality (LC 50 ) and reproduction (EC 50 ) toxicity data for standard soil test species (i.e.Lumbricidae, Collembola, Enchytraeidae, Acari) were obtained from the US Environmental Protection Agency ECOTOX database (http://cfpub.epa.gov/ecotox/), and data sets were constructed by Frampton et al. (2006) and González-Alcaraz et al. (2020).Additionally, a systematic literature search was performed on the Web of Science.Search terms and data requirements are shown in Figure 1.Relevant chemical characteristics (i.e., pK a , log K ow , DT 50 , Henry's constant) were obtained from the Pesticide Properties Database (http://sitem.herts.ac.uk/ aeru/ppdb/; Lewis et al., 2016) or from the PubChem database (https://pubchem.ncbi.nlm.nih.gov/;Kim et al., 2021).When chemical characteristics were not available in these databases, they were obtained through a literature search or estimated using Estimation Program Interface (EPI) Suite (US Environmental Protection Agency, 2021).The origin of each of the chemical characteristics used in this study is detailed in Supporting Information: Table S1.
The soil invertebrate toxicity data were extracted to a Microsoft Excel® (Microsoft) database when they met the following criteria: 1.The test chemical was organic (metals and organometal compounds were excluded); 2. The toxicity tests followed clearly described guidelines (i.e., OECD, ISO).If studies (partly) deviated from the guidelines to assess the influence of an additional parameter (i.e., temperature, organism life-stage, soil moisture content), only the toxicity data obtained by following the guidelines were included; 3. The studies clearly mentioned the soil OM (or organic carbon [OC]) content and the soil pH.Soil OC content was recalculated to OM content using the formula: OM (%) = OC (%) × 1.72, which is based on the assumption that OM contains 58% carbon (Sprengel, 1826; see also Pribyl, 2010); 4. The studies determined chemical toxicity to a species in multiple soils with varying soil OM content.This was done to ensure similar conditions in the toxicity tests (i.e., culture, handling, chemical).Consequently, data from studies using only one soil were excluded from the data set.

Data analysis
The constructed database contained toxicity data for many different chemicals and species.Most chemicals were tested on one species in a limited number of soils, and only a few chemicals were tested on multiple species in multiple soils (Figure 2).Based on this observation, it was decided to analyze the data using two approaches.
For the first approach (hereafter referred to as the "chemical-specific" approach), a subset was constructed containing the toxicity data for chemicals, which were tested on ≥2 different species and in ≥4 soils (with varying OM content) per species.In this way, the influence of OM content on the toxicity of individual chemicals could be directly assessed, and differences between species could be identified.
The subset for the second approach (hereafter named "grouped approach") contained the remaining toxicity data.As this subset consisted of many different chemicals and species, with large differences in toxicity endpoints, the data had to be "normalized" to facilitate comparisons of toxicity values for the different chemicals.This approach enabled the study to make use of the limited scope of many of the data from individual studies and to compare the data to the assumed 1:1 linear relationship between toxicity and soil OM currently applied under the correction factor of 2. For FIGURE 2 Overview of the number of toxicity data (A = LC 50 , B = EC 50 ) for each chemical and taxonomic class.The dashed line shows the minimal amount of toxicity data needed to be considered for the "chemical-specific" approach used as [Soil 1], so that ratio OM was always a value >1.For example, if the toxicity of a chemical was assessed in two soils containing 10% and 5% OM, and the corresponding LC 50 values were 5 and 2 mg/kg, the ratio OM content would be 2 and the LC 50 ratio would be 2.5.When studies used more than two soils, the ratios were calculated for each pair of soils (i.e., two soils = 1 ratio, three soils = 3 ratios, four soils = 6 ratios, five soils = 10 ratios, etc.).When OM contents of two soils were equal (i.e., ratio OM = 1), the data were excluded from the data set, as in that case, differences in toxicity would be caused by other soil properties, which were outside the scope of this study.The outcome of this decision is that the correlations reported in this study will likely be stronger than if these data were included, through a reduction in variability of toxicity values at a given OM content.However, it enabled this study to focus solely on toxicity-OM relationships in the absence of sufficient data to examine the impacts of other soil properties.Chemicals were grouped based on their respective lipophilicity, indicated by their octanol-water partitioning coefficient.The three groups were nonlipophilic (log K ow < 2), lipophilic (log K ow = 2-5), and highly lipophilic (log K ow > 5).The separation at log K ow = 2 is following the EFSA classification of lipophilic chemicals, while that at log K ow = 5 is based on previous research suggesting that the correction factor of 2 might not be applicable to highly lipophilic chemicals, as pore water may not be the primary route of exposure (Belfroid et al., 1995).Additionally, species were clustered in taxonomic groups (i.e., Lumbricidae, Enchytraeidae, Collembola, Acari), based on the assumption that species in the same taxonomic group have the same route of exposure to chemicals.
For both subsets, linear regression by least-squares approximation was used to quantify the relationship between species sensitivity and soil OM content, fitting the equation: (3) where a is the slope and b is the intercept of the regression.
Positive correlation slopes between (ratio) LC 50 or EC 50 and (ratio) soil OM content would mean that the toxicity of (grouped) chemicals decreased with increasing soil OM content, while negative correlation slopes would indicate the opposite effect.
To determine if the influence of OM content on toxicity differed significantly between species in the chemicalspecific approach, the y-intercepts (inherent toxicity of a chemical) and slopes of regressions (the degree to which toxicity is influenced by OM content) were compared by analysis of covariance, with species as the independent variable, LC 50 or EC 50 as the dependent variable, and soil OM content as the covariant.
All data manipulations were done in Microsoft Excel®, while data analyses were performed using RStudio version 1.2.1335 (RStudio Team, 2020) with R version 3.6.1 (R Core Team, 2019).

Data availability
A total of 446 toxicity data points were present in the final database, which originated from 39 studies and represented 17 soil invertebrate species and 45 organic chemicals.LC 50 data were more abundant than EC 50 reproduction data, with 249 and 197 toxicity values, respectively.For the LC 50 data, Lumbricidae were the most frequently tested organisms, and data mostly originated from the species Eisenia fetida sensu lato (E.fetida and Eisenia andrei) and Lumbricus rubellus (Table 1).Collembola were tested most often in chronic reproduction tests, with F. candida being the most tested species.Pesticides (i.e., insecticides, herbicides, fungicides) were the most investigated chemicals in both the acute and chronic toxicity tests, with insecticides being the most frequently tested chemical for both endpoints (Table 2).The effects of chlorinated chemicals such as chlorophenols, chlorobenzenes, and chloroanilines on survival and reproduction were also well studied.
The full data set was divided into two subsets (Figure 1).The first subset was used in the "chemical-specific" approach.The data set contained 80 LC 50 values and 49 EC 50 values.In total, only seven chemicals were tested on ≥2 species in ≥4 soils.All seven chemicals could be compared for the acute toxicity tests and only one chemical for the chronic toxicity tests (Supporting Information: Table S2).
The second subset contained the toxicity data that was used for the "grouped approach" and consisted of 169 LC 50 and 148 EC 50 values, for which 142 and 152 ratios could be calculated, respectively.Due to a lack of toxicity data, not enough ratios could be calculated for any taxonomic group for highly lipophilic chemicals, for Acari for nonlipophilic chemicals, and for lipophilic chemicals for Enchytraeids (Table 3).As such, these data were excluded from the data analysis.In the end, 147 LC 50 and 131 EC 50 values remained, for which 144 and 136 ratios could be calculated for mortality and reproduction tests, respectively (Table 3, Supporting Information: Tables S3 and S4).

Chemical-specific approach
For the "chemical-specific approach," toxicity data were most abundant for the herbicide phenmedipham (log K ow = 2.70), with multiple LC 50 and EC 50 values being available for the springtail F. candida and the enchytraeids Enchytraeus albidus, Enchytraeus crypticus, and Enchytraeus luxuriosus.The data were originated from studies by Amorim, Römbke, Scheffczyk, Nogueira et al. ( 2005), Amorim, Römbke, Scheffczyk, Nogueira, and Soares (2005), Domene et al. (2012), andChelinho et al. (2014), in which phenmedipham toxicity was assessed in OECD artificial soils containing 10% peat and in multiple European soils.The regressions showed a good correlation between LC 50 values and OM content for F. candida with r 2 = 0.84, while the correlations were poorer for E. albidus and E. luxuriosus with r 2 = 0.44 and r 2 = 0.50, respectively (Figure 3A).It is important to note that for F. candida the steepness of the LC 50 regression line is greatly influenced by the high LC 50 value obtained in OECD artificial soil containing 10% peat (8.0%OM).If this soil is excluded from the data analysis, the regression line is almost flat.No significant differences in regression slopes and in intercepts were found between the two enchytraeid species, suggesting that these species have similar sensitivities to phenmedipham.Therefore, the data were grouped to increase the sample size when comparing the data to those for F. candida.Significant differences in y-intercepts were observed between the enchytraeids and F. candida (F = 6.26, p = 0.02).The slope of the regression was flatter for F. candida than for the enchytraeids, which indicates a different relationship between phenmedipham toxicity and soil OM content.However, the difference in steepness of the slopes was not significant (F = 2.24, p = 0.16), which might suggest that the OM-toxicity relationship does not differ between taxa.
For the reproduction data, regression analyses for individual enchytraeid species showed decent correlations for E. albidus (r 2 = 0.41) and E. crypticus (r 2 = 0.51) but a poor correlation for E. luxuriosus (r 2 = 0.08) (Figure 3B).This suggests that toxicity−OM relationships can also differ within certain taxonomic groups.However, the regression slopes and the intercepts were not significantly different, and as such, the enchytraeid species data were grouped to

Oppia nitens 1 2
Platynothrus peltifer 1 2 increase the sample size when comparing the data to those for F. candida.Toxicity of phenmedipham to F. candida was assessed twice in an OECD artificial soil containing 10% peat (5.8% OM and 8.0% OM), but toxicity values differed by a factor of approximately 5, with reported EC 50 values of 8.04 and 39.2 mg/kg, respectively.As differences were so high between the two soils, the highest value, which fell well outside the expected range from the rest of the data, was considered an outlier and excluded from the data analysis, although the data point is included in the graph.Obtained correlations for reproduction data were poorer than those obtained from the mortality toxicity data, with r 2 = 0.24 and r 2 = 0.36 for the enchytraeids and F. candida, respectively (Figure 3B).The regression slopes for the enchytraeids and F. candida data differed in a similar manner as for the mortality data.The slope of the regression was significantly flatter for F. candida than for the enchytraeids (F = 5.64, p = 0.02), which suggests a different influence of soil OM content on phenmedipham toxicity between the two taxonomic groups.Additionally, the species significantly differed in their sensitivity to phenmedipham (F = 37.41, p < 0.05), with F. candida being the more sensitive species.To identify differences in the effect of OM content on phenmedipham toxicity endpoints, taxon specific regression slopes were compared between the two endpoints.For both taxa, the slopes of the regressions were significantly steeper for the mortality data than for the reproduction data (enchytraeids: F = 8.67, p < 0.001; F. candida: F = 16.18,p < 0.001).Taking into account that the best correlations were obtained for the mortality data, it seems that differences in soil OM content may be more relevant when assessing mortality than reproduction.
Relationships between 14-day LC 50 values and OM content for several chlorophenols, a chlorobenzene, and a chloroaniline were obtained from studies of Van Gestel andMa (1988, 1990) and Van Gestel and Van Dis (1988) (Figure 4).The species tested were E. andrei and L. rubellus, both belonging to the Lumbricidae.For chlorophenols, the regression analyses resulted in good correlations of r 2 = 0.74-0.95for E. andrei and r 2 = 0.44-0.97for L. rubellus (Figure 4A-D).Species significantly differed in their sensitivity to all chlorophenols, with E. andrei being more sensitive than L. rubellus.No significant differences were found between the slopes of the regressions of LC 50 against soil OM content for any of the chlorophenols, indicating that the influence of OM content on toxicity was similar.The only exception was 2,4,5trichlorophenol, for which significant differences were found in both the intercepts (F = 211.57,p < 0.01) and the slopes (F = 83.44,p < 0.01) (Figure 4B).For 1,2,3-trichlorobenzene and 2,4-dichloroaniline, regressions resulted in correlation coefficients of r 2 = 0.83 and r 2 = 0.98 for E. andrei and r 2 = 0.96 and r 2 = 0.80 for L. rubellus, respectively (Figure 4E,F).The slopes and intercepts were similar, showing that both species had similar sensitivities to the chemicals and that soil OM content influenced toxicity in a similar fashion.In general, no large differences were observed in the influence of soil OM content on the toxicity between the two Lumbricidae species.For both species, higher soil OM content resulted in significantly increased LC 50 values.

Grouped approach
The relationships between the ratio LC 50 /EC 50 data for Lumbricidae, Collembola, and Enchytraeidae and OM content ratios for nonlipophilic chemicals (log K ow < 2) showed that, for both the mortality and the reproduction data, variations were large and correlations were poor for all taxonomic groups (r 2 < 0.2) (Figure 5).An exception was observed in the Enchytraeidae EC 50 data, which showed a good correlation between the EC 50 ratio and OM content ratio (Figure 5F; r 2 = 0.88).In this case, the chemicals for which the ratios were calculated were RDX (log K ow = 0.87; 1 ratio) and TNT (log K ow = 1.60; 10 ratios), both explosives.The observed good correlation is possibly caused by the relatively high lipophilicity of TNT.However, the Enchytraeidae LC 50 data also included values for TNT toxicity determined in the same soils, but only poor correlations were obtained (Figure 5E; r 2 = 0.18).As such, this observation remains currently unexplained.
In the Lumbricidae mortality data, several high LC 50 ratios/ low OM ratios were found (Figure 5A).These points were obtained for the fungicides benomyl (log K ow = 1.40) and carbendazim (log K ow = 1.48), which is one of the main metabolites of benomyl.The toxicity of these chemicals apparently is greatly reduced at high soil OM content.Two high EC 50 ratios/low OM content ratios were also observed among the Collembola reproduction data, which were both for the insecticide imidacloprid (Figure 5D).One of these data points was calculated from toxicity data for an OECD artificial soil containing 10% peat and a LUFA 2.2 soil.The other high EC 50 /low OM content ratio for imidacloprid was calculated from toxicity data generated in an oxisol and an entisol.
The relationships between the ratio LC 50 /EC 50 data and soil OM ratios for Lumbricidae, Collembola, and Acari for lipophilic chemicals (log K ow = 2-5) resembled those for the nonlipophilic chemicals (Figure 6).Regression analyses mostly resulted in poor correlations for both the mortality and the reproduction data (r 2 < 0.20).The only exception was for the Lumbricidae mortality data (Figure 6A; r 2 = 0.36).The LC 50 data contains several OM content ratios of 2.19, with corresponding LC 50 ratios ranging between 1.00 and 3.34.These data originate from several studies by the same author, in which the toxicity of a multitude of chemicals was assessed in the same two soils.This is also the case for the cluster observed in the Collembola data at OM ratios of 2.63 (Figure 6C,D).These data originated from the same study in which the effect of nine chemicals on Collembola survival and reproduction was assessed in OECD 10% and LUFA 2.2 soils.These data illustrate the variability within the OM-toxicity relationship even within the same studies.

Differences between species
One of the aims of the current study was to investigate if the relationships between toxicity and soil OM content   3 for the toxicity data used to prepare this figure differed between species.The exposure of soil invertebrates to chemicals occurs through contact with pore water, ingestion of soil particles, and inhalation of air present in the soil pores.The relative importance of each of these uptake routes is determined by morphological, physiological, and behavioral factors (Peijnenburg et al., 2012).In general, invertebrates can be subdivided into two groups: soft-bodied and hard-bodied.Hard-bodied organisms have evolved special organs for the assimilation of oxygen and water, while in soft-bodied biota, these processes are regulated through uptake via the skin.Contaminants and nutrients may also be taken up via these distinct exposure routes, while uptake of contaminants via food is possible for all biota.These different exposure routes may result in different toxicity−OM content relationships.Additionally, body size may be an important factor, as chemical uptake may be faster in smaller than in larger invertebrates due to the larger surface-to-volume ratio of the former ones (Dalhoff et al., 2020).
The results from the current study show some indications of differences in the toxicity−OM content relationship between hard-bodied and soft-bodied soil invertebrates.
The relationships obtained for the acute toxicity to E. andrei and L. rubellus of chlorophenols, chloroanilines, and chlorobenzenes were similar in most cases.Both of these species are soft-bodied, belong to the phylum Lumbricidae, and have a similar ecology.Furthermore, the acute and chronic toxicity−OM content relationships for soft-bodied enchytraeids and hard-bodied F. candida exposed to phenmedipham differed, although only significantly for the chronic toxicity data.It has previously been suggested that for collembolans the uptake of chemicals is associated with solid soil phases, while for soft-bodied species, it is more influenced by pore water (Vijver et al., 2001).Enchytraeids are soft-bodied invertebrates and are in constant contact with the pore water, more so than F. candida, which is a hard-bodied invertebrate.As such, the reduction of the concentrations in the pore water at higher soil OM content may have a stronger influence on the toxicity of chemicals to enchytraeids than to F. candida.
Although no comparisons of the toxicity−OM content relationships for Eisenia sensu lato and F. candida could be made in the current study due to insufficient data, the results obtained for phenmedipham show that differences between enchytraeids and F. candida do exist.As enchytraeids and earthworms are both soft-bodied and enchytraeids are the closest relatives to the earthworms (Erséus & Källersjö, 2004), it seems likely that uptake routes for these species are similar.Thus, differences in the influence of OM content on toxicity could also be present between earthworms and F. candida.However, more research is needed to verify this, especially since earthworms (E.fetida sensu lato) and F. candida are the most used standard test organisms in soil risk assessment.

Differences between endpoints
Another aim of this study was to investigate if the relationships between toxicity and soil OM content were similar for different toxicity endpoints.The results from the "chemical specific" approach, in which individual chemicals were analyzed, only had mortality and reproduction data available for phenmedipham.For this chemical, soil OM content had a larger influence on LC 50 than EC 50 values for Enchytraeus sp. and F. candida, indicated by higher correlation coefficients and steeper slopes of the regression lines.For F. candida, the steepness of the LC 50 regression line is greatly influenced by the high LC 50 value obtained in OECD artificial soil containing 10% peat.If this soil is excluded from the data analysis, the regression line is almost flat, which is similar to the regression line for the EC 50 data.Thus, these results suggest that the influence of soil OM on toxicity endpoints may also differ between species.For enchytraeids, soil OM influences acute toxicity more than chronic toxicity, while for F. candida, soil OM content plays a smaller role in both acute and chronic toxicity.For the second approach, in which chemicals were grouped based on their lipophilicity, no clear differences were identified between the two endpoints, which is most likely due to the weak correlations obtained in this method.3 for the toxicity data used to prepare this figure Correlation coefficients of the regression lines were lower for the EC 50 than for the LC 50 data, which suggests that other soil properties are more important for reproduction than for mortality.In the case of phenmedipham, the higher LC 50 values observed in soils containing higher OM content may be a direct result of increased sorption to soil OM, thus resulting in lower bioavailability.In contrast, reproduction is a more sensitive endpoint, which may also be strongly affected by other soil parameters.For instance, Van Gestel (1992) reported that the reproduction of E. andrei was optimal at a soil pH of 5-6.At higher and lower pH, the number of juveniles per worm per week was significantly reduced.Similarly, Greenslade and Vaughan (2003) reported that soils with a pH between 5.4 and 6.6 were ideal for the reproduction of F. candida.The number of offspring decreased by about 50% at a pH of 3.5, while no reproduction occurred in soils with a pH of 8. Jänsch et al. (2006) reported that the ideal soil pH for enchytraeid reproduction was between 5.8 and 7.0.Soil OM content may therefore be of less importance in affecting (the toxicity of chemicals to) reproduction due to the confounding effects of other edaphic properties like, for example, soil pH.

Influence of other soil properties
In the current study, only the influence of soil OM content on the toxicity of chemicals was investigated.The soils used in the studies from which the toxicity data was obtained often also differed in other properties besides OM content, such as clay and silt content, pH, water holding capacity, cation exchange capacity (CEC), OM type, and C/N ratio.These properties may influence species sensitivity to a chemical, either through direct test species performance (Domene et al., 2012) or through effects on chemical bioavailability (Delle Site, 2001).Most likely, some of the observed variations in toxicity endpoints is influenced by differences in these soil properties.This seems to be the case for imidacloprid in the "grouped approach," where the OM content of the oxisol and entisol differed by a factor of 2, while the clay content differed by a factor of 15, and the toxicity differed by a factor of 18 (Figure 5D).In addition, the OM content of an OECD 10% and LUFA 2.2 soil differed by less than a factor of three, but the toxicity differed by a factor of 15.Overall, quantifying the influence of these soil properties together is difficult because of the complex interactions between them, which are often intercorrelated.For example, CEC is partly determined by soil clay and OM content, as they provide cation-exchange sites.However, the CEC is also influenced by pH as the cation-exchange sites on soil OM are pH-dependent (Dayton et al., 2006).Additionally, in many studies, these soil properties are not reported, further complicating an in-depth analysis of their influence on toxicity.In the current study, quantifying the influence of these properties on toxicity therefore was not possible; however, further study in this area is warranted.

Implications for risk assessment
The EFSA recommends the application of the correction factor of two to toxicity endpoints when chemicals are lipophilic (i.e., log K ow > 2) to correct for the difference in soil OM content between artificial (10% sphagnum peat) and agricultural soils (<5% OM) (EFSA, 2017).This assumes that the bioavailability (and thus toxicity) of lipophilic chemicals is inversely related to soil OM content and that a 1:1 linear decrease in toxicity would be expected as OM content increases.Following this assumption, if the OM content between two soils differs by a factor two, the difference in toxicity should also be a factor two (for lipophilic chemicals), and data points in the "grouped approach" should fall along the 1:1 linear relationship line in Figures 5 and 6.
The results obtained in the "grouped approach" do not support this assumption.For both lipophilic and nonlipophilic chemicals, no clear (linear) relationship between toxicity and OM content could be obtained and data points were scattered both above and below the 1:1 linear relationship line (Figures 5 and 6).This was expected for nonlipophilic chemicals, as they are less likely to sorb to the OM fraction in soils.However, for the lipophilic chemicals, the results were unexpected, as the "chemical-specific" approach showed several good correlations between soil OM and toxicity.Hence, it seems that the influence of OM content on toxicity cannot be quantified when chemicals are grouped solely by their lipophilicity.This observation is further supported by the chronic EC 50 ratio data for Enchytraeidae in the grouped approach, which showed only a clear positive linear relationship between the ratio EC 50 and OM content ratio (r 2 = 0.88) (Figure 5F).In this case, 10 of the 11 ratios were obtained for a single chemical, namely, TNT (log K ow = 1.6).
Of course, chemicals differ in other chemical characteristics besides lipophilicity, which plays a role in the soil−water partitioning and therefore bioavailability to soil organisms.These chemical properties involve, among others, water solubility, acidity, molecular size, polarity, and polarizability (Von Oepen et al., 1991).The importance of these chemical properties in determining bioavailability can in turn be influenced by soil properties other than just OM content.For instance, for ionizable chemicals such as certain chlorophenols, the soil pH can determine if the chemical is present in the (an) ionic or the phenolic form.(An)ions generally adsorb poorly onto OM (although clay may be important, especially for cations), while sorption of the phenolic form is determined mostly by lipophilicity (Van Gestel, 1992).It might perhaps help to make comparisons between soils if sorption coefficients would be available for the different test soils, but unfortunately, this was rarely the case.
In the current study, chemicals were considered lipophilic if their log K ow ranged between 2 and 5, which is an arbitrary decision, although based on current approaches in pesticide regulation (EFSA, 2017).In this category, chemicals with higher log K ow values can be expected to exhibit stronger sorption to soil OM than ones with a lower log K ow .At strong binding to the soil, soil passing the gut may become an additional route of exposure (Belfroid et al., 1995).Consequently, differences in uptake (and consequent toxicity) can be smaller for chemicals with log K ow s of, for example, 1.9 and 2.1 than for log K ow s of, for example, 2.1 and 4.9.
In general, it seems that toxicity−OM content relationships differ between chemicals, species, and endpoints.The correction factor of two was developed based on 14-day earthworm mortality data.The results of the current research suggest that the influence of OM on pesticide toxicity is the largest for mortality in soft-bodied invertebrates and is less important for reproduction and hard-bodied invertebrates.Since European risk assessment now uses more species than just earthworms and focuses on reproduction instead of mortality, the application of the correction factor of two for lipophilic pesticides may not be justified in all cases.Finally, the results of the current research also showed that the toxicity of chemicals that are currently considered nonlipophilic (i.e., log K ow < 2) can also be influenced by soil OM content.

Data limitations
In Europe, the hazard of pesticides to soil invertebrates was assessed by investigating the toxicity to only one species, most frequently Eisenia sensu lato.For a decade, two additional test species, the arthropods F. candida and H. aculeifer, have to be tested if the pesticide reaches the soil (EFSA, 2017).The generated toxicity data are derived in the context of risk assessment, and as such, the tests primarily follow standard regulatory testing protocols, which only use a standardized artificial soil with sphagnum peat as the type of OM, at a content of 10% or 5%.Consequently, pesticide toxicity data are only very rarely publicly available for more than one species (earthworms) and they are only very rarely assessed in more than one soil type, hampering the ability to investigate the influence of soil properties on species sensitivities.This was apparent in the current study, as only 446 toxicity values were included in the final database.Of these toxicity data, only 129 of the 446 values were for chemicals tested on two or more species in four or more soils.As such, direct comparisons of the influence of OM content on the toxicity between species were only possible for seven chemicals, and only the herbicide phenmedipham had acute and chronic toxicity data available for several different species in more than four soils.Additionally, earthworms and springtails were overrepresented in the mortality and reproduction data, respectively, while mites were underrepresented in both.Although mites seem relatively nonsensitive to pesticides (Daam et al., 2011;Frampton et al., 2006), it is important to understand the influence of soil properties on these species, as their uptake routes of chemicals are still poorly understood (Huguier et al., 2015), and to ensure ecological relevance of the risk assessment.Overall, due to this lack of data, the extent to which the influence of soil OM content on species sensitivity could be investigated was limited.
To use the obtained toxicity data for chemicals with few toxicity data available, a grouped approach was applied in which ratios between two soils were calculated.Species were grouped based on their taxon, and chemicals were grouped based on their lipophilicity.The division in taxonomic groups assumes that the different species within a family have a similar relationship between toxicity and OM content, which does not necessarily have to be the case.This is especially true for Collembola, which are a highly diverse group with species being found all over the globe in different habitats.However, most toxicity data for Collembola were for F. candida (96 of 101 ratios).The other Collembola species for which toxicity data were available were Folsomia fimetaria and Onychiurus folsomi.These species do not differ that much from F. candida, although F. candida is parthenogenic, while the other two species reproduce sexually.So, in the case of chemicals acting specifically on reproduction (e.g., sex hormone-disrupting chemicals), these species may show a difference in response, but otherwise these species were expected to respond similarly to differences in soil OM content.
The relationship between toxicity and soil OM content may also differ between Lumbricidae species, which can be classified into three ecological groups: epigeic, anecic, and endogeic (Bouché, 1975).Epigeic species live on the soil surface and feed primarily on litter.Anecic worms create and reside in permanent vertical burrows up to 3 m deep in the soil and also feed primarily on the litter layer.Endogeic worms create and live in horizontal burrows in the upper soil layer (25-40 cm), feeding on soil OM and rarely coming up to the soil surface.Most likely, species differing in their ecology will also differ in their physiology and interaction with the soil, and, as a consequence, the route of exposure to chemicals may also be different between ecological groups.For instance, the uptake of chemicals by the gut may be more important for endogeic species than for epigeic and anecic species, as they ingest more soil to feed on soil OM.This may especially be important for highly lipophilic chemicals in soils containing high OM content.Due to these differences in ecology, grouping earthworm species together may not always be appropriate.In the current study, the "grouped approach" Lumbricidae toxicity data consisted of eight and two species for LC 50 and EC 50 , respectively.For the EC 50 data, these species were E. fetida and E. andrei, which are very closely related and have similar physiological properties and exposure routes to chemicals, and therefore, grouping them was deemed justified.The Lumbricidae LC 50 data for lipophilic chemicals originated from four species, E. fetida, E. andrei, Lumbricus terrestris, and L. rubellus, which all feed on litter on the soil surface, and therefore, similar routes of exposure were also expected.Finally, the nonlipophilic chemicals were tested for eight Lumbricidae species, which were the previously mentioned ones, plus Allolobophora chlorotica, Aporrectodea caliginosa, Aporrectodea rosea, and Aporrectodea tuberculata.The latter four species are all endogeic, living in and feeding on soil, thereby ingesting relatively large quantities of soil.Thus, in the Lumbricidae nonlipophilic chemical data analysis, epigeic, anecic, and epigeic species were grouped together.However, as the chemicals were not lipophilic, ingestion was not expected to be an important route of exposure.Thus, overall, differences in routes of exposure (i.e., through ingestion) were not expected in the data analyses, and grouping species based on their taxon was considered to be justified.
To address these limitations and further improve our understanding of the influence of soil OM content on pesticide toxicity to soil invertebrates, acute and chronic toxicity data should be generated for pesticides ranging in their lipophilicity.These pesticides should be tested on two or more species in at least four similar soils with different OM content.In this way, more direct comparisons between toxicity and soil OM content can be made while also highlighting the influence of chemical characteristics on the toxicity−OM content relationship.A thorough analysis of a dataset constructed in this manner would allow for a more scientifically sound underpinning of correction factors, resulting in a more ecologically relevant risk assessment.

Artificial soils-A need for peat standardization
One of the problems with artificial soils is that, although they consist of the same components and are intended to enable standardization between studies, their soil properties can still vary greatly.To illustrate this, Hofman et al. (2009) compared the properties and sorption capacity of artificial soils collected from 20 European laboratories.They found large variations in the soil properties and cadmium sorption to these artificial soils differed by over one order of magnitude.Additionally, the OC content of the soils in their study varied from 1.4% to 6%, even though all soils were prepared with 10% sphagnum peat.They concluded that sphagnum peat is the most problematic component of artificial soils, as its properties (i.e., OC content, pH, microbial activity) can differ greatly between countries and suppliers.Furthermore, the manipulation of the peat prior to the construction of the artificial soil (i.e., drying, grinding, sieving) is not fully standardized by the OECD test guidelines, which may influence its sorption capacity for organic chemicals.
The abovementioned problems were also observed in the toxicity data collected for this study.In total, 24 studies used artificial soil containing 10% peat, of which 17 measured the actual soil OM content.The reported OM values differed by a factor of 5 and ranged from 2.2% to 11.8% (Table 4), which is similar to the range described by Hofman et al. (2009).The difference in soil OM content also resulted in differences in toxicity.For example, Domene et al. (2012) and Amorim, Römbke, Scheffczyk, Nogueira et al. (2005) exposed F. candida to phenmedipham in artificial soils containing 10% peat (5.8% and 8.0% OM) but report EC 50 values differing by a factor of 5 (8.04 and 39.2 mg/kg).The EC 50 value of 8.04 mg/ kg corresponded well with the values obtained in natural soils, but the EC 50 of 39.2 mg/kg was significantly higher than the other values, even when considering the higher OM content.It is unclear what the exact cause is of these differences, but they may be due to differences in the sphagnum peat used to construct their artificial soils.Unfortunately, both studies did not report on the origin of their sphagnum peat nor how the peat was handled prior to use.
An additional problem of artificial soils is that the type of OM and clay is different from natural soils.While sphagnum peat is derived from slowly decayed and decomposed freshwater biomass, natural soils contain (not fully decomposed) OM from the terrestrial environment.As such, the chemical and physical properties of sphagnum peat differ from the OM found in natural soils.Additionally, kaolin clay has a low adsorption capacity when compared to other naturally occurring clay types, possibly leading to higher bioavailability of certain chemicals.Previous studies have shown that OM quality and chemical properties play an important role in changing the sorption behavior of lipophilic organic chemicals (Belfroid et al., 1996;Hofman et al., 2008;Peters et al., 2007).In a study performed by Vlčková and Hofman (2012), the bioaccumulation of several lipophilic chemicals in E. fetida was assessed in artificial and natural soils containing similar OM content.They found that the bioavailability of the chemicals was mostly underestimated in artificial soils when compared with natural soils.Vlčková and Hofman (2012) concluded that humic acids and humin were the most dominant fractions of soil OM contributing to total soil sorption, as the artificial soils contained two to four times higher humic acid/fulvic acids ratios than the natural soils, which is in accordance with Liu et al. (2010).
Despite the calls from Hofman et al. (2009) for the standardization of the peat component in artificial soils over a decade ago, no such standardization has taken place.The current study endorses the earlier call for further standardization.As of right now, studies should include more information on the type of peat used in their soils (i.e., brand, manipulations [drying, sieving]).Ideally, studies would also include the amount of OM or OC in their artificial soil.Soil OM content can be measured through simple loss on ignition techniques.The technique used to measure OM or OC content should be described clearly, as factors such as furnace type, sample mass, duration, temperature of ignition, method of sample preparation, and type of TOC apparatus can all influence the accuracy of the measurement (Hoogsteen et al., 2015).Furthermore, the peat component should be further explored regarding its ecological impacts.Suitable alternatives have been identified, such as coir (De Silva & van Gestel, 2009), which enable the successful conduct of earthworm toxicity testing.Further research could consider identifying whether such more sustainable alternatives provide a more consistent quantity of OM or OC into the artificial substrate.

CONCLUSION
This study showed no clear relationship between chemical toxicity and soil OM content when chemicals were grouped by lipophilicity.The chemicals also differed in other characteristics (i.e., lipophilicity, solubility, polarity, etc.), which may explain the absence of a clear relationship.Overall, the results indicate that grouping chemicals based on their lipophilicity is not appropriate, as toxicity-OM relationships appear to be chemical-specific.Indeed, several clear toxicity−OM content relationships were obtained for individual chemicals.For these chemicals, toxicity decreased with increasing soil OM content, and the influence was stronger for mortality than for reproduction, suggesting that consideration of the OMtoxicity relationship for individual endpoints is warranted.Additionally, it seems that soil OM content influences toxicity more for soft-bodied (i.e., earthworms and enchytraeids) than for hard-bodied invertebrates (i.e., springtails and mites), which is likely due to differences in exposure routes.
The current research shows that quantifying the influence of soil OM content on the toxicity of organic chemicals to soil invertebrates is difficult, as toxicity is rarely assessed in more than one soil causing a lack of relevant toxicity data.Additionally, soil properties are often intercorrelated and have complex interactions with each other, hampering the establishment of cause-and-effect relationships between toxicity and OM content.The measured OM content in standardized OECD artificial soils prepared with 10% peat, reported in the literature, differed by a factor of 5.The use of different types of sphagnum peat may have a big impact on toxicity results, weakening comparability among studies.Hence, there is an urgent need to better define the peat component in artificial soils used in toxicity studies.
Ultimately, the current study shows that the relationship between toxicity and soil OM content is highly complex.For a better understanding, more research is needed to quantify the toxicity−OM content relationships for different species, endpoints, and chemicals.This will allow for a more scientifically sound underpinning of correction factors, resulting in a more ecologically relevant risk assessment.

FIGURE 5
FIGURE 5 Ratios of LC 50 (left)/EC 50 (right) versus ratios of soil organic matter (OM) contents for the toxicity of nonlipophilic chemicals (log K ow < 2) to soil invertebrates in different soil types.Markers represent the ratio between toxicity values obtained in soils with different OM content, colored lines are regression lines.Shown are data for Lumbricidae (A, B), Collembola (C, D), and Enchytraeidae (E, F).The dotted black line shows the assumed 1:1 linear relationship for the OM correction factor.See Table3for the toxicity data used to prepare this figure

FIGURE 6
FIGURE 6 Ratios of LC 50 (left)/EC 50 (right) versus ratios of soil organic matter (OM) contents for the toxicity of lipophilic chemicals (log K ow = 2-5) to soil invertebrates in different soil types.Shown are data for Lumbricidae (A, B), Collembola (C, D), andAcari (E, F).The dotted black line shows the assumed 1:1 linear relationship for the OM correction factor.See Table3for the toxicity data used to prepare this figure Overview of the toxicity data for chemical effects on soil invertebrates used in the present study, sorted by taxonomic groups Integr Environ Assess Manag 2023:1457-1472 © 2023 The Authors wileyonlinelibrary.com/journal/ieamTABLE 1

TABLE 2
Overview of the toxicity data for organic chemicals on soil invertebrates used in the present study, sorted by chemical type Integr Environ Assess Manag 2023:1457-1472 © 2023 The Authors DOI: 10.1002/ieam.4770Abbreviation: n, number of toxicity data available.

TABLE 3
Overview of the toxicity data on soil invertebrates used in the "grouped approach" for assessing the effect of soil organic matter content on toxicity, sorted by toxicity endpoint, chemical lipophilicity, and taxonomic group Integr Environ Assess Manag 2023:1457-1472 © 2023 The Authors wileyonlinelibrary.com/journal/ieam Integr Environ Assess Manag 2023:1457-1472 © 2023 The Authors DOI: 10.1002/ieam.4770TABLE 4 Reported soil organic matter (OM) content (%) in studies using OECD artificial soil containing 10% sphagnum peat