Comparison of Species Sensitivity Distributions for Sediment-Associated Nonionic Organic Chemicals Through Equilibrium Partitioning Theory and Spiked-Sediment Toxicity Tests with Invertebrates.

Equilibrium partitioning (EqP) theory and spiked-sediment toxicity tests are useful methods to develop sediment quality benchmarks. However, neither approach has been directly compared based on species sensitivity distributions (SSDs) to date. In the present study, we compared SSDs for 10 nonionic hydrophobic chemicals (e.g., pyrethroid insecticides, other insecticides, and polycyclic aromatic hydrocarbons) based on 10-14-day spiked-sediment toxicity test data with those based on EqP theory using acute water-only tests. Because the exposure periods were different between the two tests, effective concentrations (i.e., median effective/lethal concentration) were corrected to compare SSDs. Accordingly, we found that hazardous concentrations for 50% and 5% of species (HC50 and HC5, respectively) differed by up to a factor of 100 and 129 between the two approaches, respectively. However, when five or more species were used for SSD estimation, their differences were reduced to a factor of 1.7 and 5.1 for HC50 and HC5, respectively, and the 95% confidence intervals of HC50 values overlapped considerably between the two approaches. These results suggest that when the number of test species is adequate, SSDs based on EqP theory and spiked-sediment tests are comparable in sediment risk assessments. Environ Toxicol Chem 2022;41:462-473.

INTRODUCTION

Although widespread sediment contamination has been recognized as an area of concern, the risks that it carries require further investigation and countermeasures. Setting sediment quality benchmarks is a fundamental step in addressing this challenge. These benchmarks are a vital step in assessing the ecological risks of contaminated sediments, identifying contaminated sites, and prioritizing chemicals of concern. Two major approaches have been used to develop sediment quality benchmarks for nonionic hydrophobic organic chemicals (HOCs): equilibrium partitioning (EqP) theory (US Environmental Protection Agency [USEPA], 2012) and spiked‐sediment toxicity tests (EFSA Panel on Plant Protection Products and Their Residues, 2015).

Equilibrium partitioning theory assumes that a nonionic chemical is at equilibrium between sediment organic carbon, interstitial water (i.e., porewater), and benthic organisms. If chemical activity in any one phase is known to be at equilibrium, the chemical activity in the other phases can be predicted (Burgess et al., 2013; Di Toro et al., 1991; Shea, 1988). Thus, if the effective concentration of a target chemical in water (e.g., micrograms per liter, median lethal concentration [LC50]) is known, the effective concentration in sediments on an organic carbon basis (e.g., micrograms per gram organic carbon) can be reasonably predicted. This can be performed by multiplying the effective concentration in water with the organic carbon–water partition coefficient (K OC; Di Toro et al., 1991). Di Toro et al. (1991) also suggested that if the sensitivity of benthic organisms to an HOC is not different from that of pelagic organisms, the sediment quality benchmarks of the nonionic HOC can be derived using water quality benchmarks (or toxicity test records with pelagic organisms) and K OC for the same chemical (Di Toro et al., 2000). Sometimes EqP theory is misunderstood as a method that does not take into account the effects of sediment ingestion as an exposure route. However, effective exposure concentration is the same regardless of exposure route if the sediment, porewater, and benthic organisms are at equilibrium (Di Toro et al., 1991); thus, EqP theory can be applied even when sediment ingestion is the dominant exposure route (Jager, 2004).

Spiked‐sediment toxicity tests, in contrast, are conducted by exposing benthic organisms to noncontaminated sediments, which are spiked with a test chemical in the laboratory. This approach provides direct evidence of the concentration–response relationship in benthic organisms for a single toxicant. Spiked‐sediment toxicity tests have been standardized for several benthic organisms, including amphipods, midges, oligochaetes, and polychaetes (Organization for Economic Co‐operation and Development, 2004, 2007, 2010; USEPA, 2000). Several endpoints have been suggested for these standardized test methods to measure the effects of contaminants, including survival, growth, and reproduction; survival after 10‐day exposure is required by regulations such as the Federal Insecticide, Fungicide and Rodenticide Act and is the most commonly reported endpoint (USEPA, 2000, 2016).

Both approaches have their own advantages and limitations. The advantage of the EqP‐based approach is that more toxicity test records are available for pelagic organisms with a variety of HOCs than for those with benthic organisms. This, in turn, is a drawback of spiked‐sediment toxicity tests, which have been developed for only a few benthic species (e.g., amphipods and midges); thus, existing test records may capture only a limited range of benthic species sensitivities (Nowell et al., 2016). In contrast, a major limitation of the EqP‐based approach is the uncertainty of K OC values, which differ depending on the composition of organics, and thus vary among sediment samples (Chefetz et al., 2000; Endo et al., 2020; Hawthorne et al., 2006). Spiked‐sediment toxicity tests have a similar limitation because toxicity values vary depending on the types of sediment used. Nonetheless, spiked‐sediment test results are considered more direct measurements than the EqP approach for quantifying exposure–effect relationships in benthic organisms (Nowell et al., 2016).

Establishing sediment quality benchmarks requires predicted‐no‐effect concentrations (PNECs) of a chemical that can typically be derived using two methods. The first is by applying an assessment factor (AF; also known as the uncertainty factor) to the lowest effective concentration available (hereafter called the AF method). The second method involves estimating the species sensitivity distribution (SSD). In the SSD approach, effective concentrations of a set of species are typically fitted to a certain statistical distribution (e.g., log‐normal and log‐logistic), and a hazardous concentration for x% of the species (HCx, typically HC5) is estimated and thus used to derive the PNEC. This SSD approach is more advantageous than the AF method because it thoroughly uses available toxicity data by explicitly expressing associated uncertainties (Belanger et al., 2017; Sorgog & Kamo, 2019). Nevertheless, the regulatory use of SSD requires toxicity data for at least five to 10 species (Belanger et al., 2017; EFSA Panel on Plant Protection Products and Their Residues, 2015), thus hampering the wide application of SSD to spiked‐sediment toxicity tests. In contrast, because the EqP approach uses toxicity data for pelagic species, for which data are more abundant than for benthic species, it has greater potential than the spiked‐sediment test to derive SSDs for a wider range of HOCs.

Considering these limitations, it remains inconclusive whether the EqP approach or spiked‐sediment test is more appropriate for the development of sediment quality benchmarks. Although such evaluation and comparison of the two approaches based on SSDs are critical, to the best of our knowledge, only few attempts have been made to date. For example, Redman et al. (2014) compared EqP theory with sediment and soil toxicity tests based on SSDs for nonpolar organic chemicals by applying a target lipid model (TLM). The TLM assumes that the critical target lipid body burden is organism‐specific but not chemical‐specific; application of the TLM is thus restricted to the baseline narcotic toxicity (Boone & Di Toro, 2019; Di Toro et al., 2000). In the present study, we compared EqP theory and spiked‐sediment tests based on SSDs for nonionic HOCs, without application of the TLM, which allows comparison of the two approaches with a wider range of modes of action. We used comprehensive ecotoxicity databases to derive SSDs. In addition, we investigated whether the relationships between SSDs estimated by the two approaches were affected by sample size (i.e., number of tested species).

MATERIALS AND METHODS

Ecotoxicity data compilation

Data on spiked‐sediment toxicity tests were collected from the Society of Environmental Toxicology and Chemistry Sediment Interest Group (formerly SEDAG) database (October 2016 updated version; Society of Environmental Toxicology and Chemistry, 2016) and the peer‐reviewed literature. The literature search was conducted using Google Scholar, and several studies were retrieved from a comprehensive review by Deneer et al. (2013). Collected data included a total of 1705 in vivo ecotoxicity test records for 49 species and 100 chemicals. Only toxicity data for HOCs with a log octanol–water partition coefficient (K OW) >3 (EFSA Panel on Plant Protection Products and Their Residues, 2015), LC50 as an endpoint, and an exposure period of 10–14 days (sometimes called subchronic tests) were used for further analyses.

Data on water‐only toxicity tests were collected from the EnviroTox database, Ver 1.3.0 (https://envirotoxdatabase.org; Connors et al., 2019), in December 2020. The EnviroTox database retrieves and curates ecotoxicity data from other databases such as the USEPA's ECOTOX Knowledgebase (https://cfpub.epa.gov/ecotox/) and the European Chemicals Agency's publicly available Registration, Evaluation, Authorisation and Restriction of Chemicals data (Connors et al., 2019). Data collected from this database included 79,585 in vivo ecotoxicity test records for 1546 species and 3989 chemicals. Because only limited toxicity data sets of water‐only lethality tests with exposure periods of 10 or more days were available in the EnviroTox database, 10‐day water‐only toxicity data (Supporting Information, Table S3) were additionally collected from the published literature to correct the effect of exposure period (see the next section for further details). Only invertebrate data were used for further analyses of water‐only tests because the test organisms in the spiked‐sediment toxicity tests were only invertebrates. In addition, only toxicity data with acute LC50 as an endpoint and for HOCs with log K OW >3 were used for further analyses. The LC50 values were based on either nominal or measured (e.g., time‐weighted average and arithmetic mean) concentrations. Both saltwater and freshwater species were included in the analyses, to examine as many chemicals with toxicity data with at least three species as possible. No systematic differences were present in mean acute effective concentrations (i.e., median effective concentration [EC50] and LC50) between freshwater and saltwater organisms, as demonstrated by previous studies (De Zwart, 2002). In addition, our investigation showed that differences in HC50 values derived from SSDs of freshwater and saltwater invertebrate species were within a factor of 10 for chemicals considered in the present study, with the exception of a factor of 21 for permethrin (Supporting Information, Figure S2).

To demonstrate the validity of the SSD estimation using the EnviroTox database, we compared the estimates of HC5 values in the present study with those reported in previous studies using only invertebrate species, without applying EqP theory (Supporting Information, Table S4).

EqP theory

In water‐only tests LC50 values (C water; expressed as milligrams per liter) were converted using K OC and Equation (1) according to EqP theory (Di Toro et al., 1991) as follows: In Equation 1, C sed,EqP is the EqP‐based organic carbon normalized sediment LC50 value (expressed as milligrams per kilogram of organic carbon), and K OC (expressed as liters per kilogram of organic carbon) of the chemical under consideration was estimated using the molecular connectivity index method (Sabljlc, 1987), implemented in the KOCWIN program of EPI Suite, Ver 4.1. Although a more elaborate EqP model that incorporates black carbon (e.g., soot and coke) as a second partitioning phase has been proposed (McGrath et al., 2019), such a model was not employed in the present study because of the unavailability of reliable black carbon–water partition coefficients for the chemicals considered in the present study.

The exposure periods for most water‐only acute toxicity tests retrieved from EnviroTox ranged from 24 to 96 h, whereas those of spiked‐sediment toxicity tests ranged from 10 to 14 days. Therefore, the converted toxicity value (i.e., C sed,EqP) was further corrected to account for this difference. An increase in chemical hydrophobicity increases the time taken to reach equilibrium; the following equations were used for the correction (Di Toro et al., 2000): In Equations 2 and 3, the subscript number in C sed,EqP (i.e., t or 10) indicates the exposure period in days in a water‐only test with a pelagic organism, and K OW was retrieved from the EnviroTox database. The coefficients in Equation 2 were retrieved from the study conducted by Di Toro et al. (2000), and those in Equation 3 were derived from the data collected in the present study. No association was found between K OW and the ratio of 96‐h to 10‐day LC50s; thus, the median (2.56) of the ratio of 96‐h to 10‐day LC50s was used for the correction (Equation 3). Note that when the 2.5% and 97.5% quantiles of the ratio were used, the resulting HC50 values were approximately two times higher and 3.6 times lower than those using the median ratio (Supporting Information, Table S5). See the Supporting Information for more details.

Data analysis

All data analyses were performed following Hiki and Iwasaki (2020), with slight modifications, using R software, Ver 4.0.5 (R Foundation for Statistical Computing, 2020), and the R package “tidyverse” (Wickham, 2017). Data visualization was performed using the R package “ggplot2” (Wickham, 2016). The R code and all the data used are available in the Supporting Information and at https://github.com/KyoHiki/EqP_vs_Spiked_sediment_test.

For nonionic HOCs, SSDs were derived using data from water‐only toxicity tests (i.e., the EqP‐based approach) and spiked‐sediment toxicity tests. The SSD for the chemical under consideration was estimated only if water‐only or sediment toxicity data for that chemical were available for three or more species. Note that the minimum of three species was below the number of species required by the regulations (Belanger et al., 2017; EFSA Panel on Plant Protection Products and Their Residues, 2015) but was adopted in the present study to examine as many chemicals as possible. In accordance with this criterion, we derived SSDs for 165 and 10 chemicals based on water‐only and sediment tests from 3219 test records (3012 water‐only and 207 sediment tests) of 409 species, respectively. All the data used in the SSD derivation are provided in Supporting Information, Tables S1 and S2. If multiple effect concentrations were available for a given combination of species, chemical, and data source (i.e., EqP or spiked‐sediment tests), their geometric means were calculated and used for analyses. The mean (i.e., HC50) and standard deviation (SD) of log‐normal SSDs were estimated from the log10‐transformed concentrations using the “mean” and “sd” functions in the R packages “base” and “stats,” respectively. The SD of SSD was calculated as the square root of the sum of the squared deviations from the mean divided by n − 1, where n is the number of species. The HC5 value was estimated using Equation 4:

The confidence intervals (CIs) of HC50 and HC5 were estimated according to Aldenberg and Jaworska (2000). The R code to calculate CIs is available in the Supporting Information. The normality of SSDs was assessed using the Shapiro‐Wilk test at an α level of 0.05, with Holm's p‐value adjustment. The assumption of normality for any SSD derived was not rejected.

In addition, as a supplementary analysis, we investigated the effects of the variation in reported sediment toxicity data, which could have been affected by differences in laboratories, experimental conditions, and/or sources of organisms, on the HC50 and HC5 estimates. For this analysis, a total of 1000 SSDs were estimated for each chemical by randomly resampling one toxicity value from each test species using the “sample_n” function in the R “dplyr” package (Supporting Information, Figure S5). The number of test species was kept the same as in the original full data set. The differences between the 97.5th and 2.5th percentile values of HC50 and HC5 in 1000 SSDs for a chemical were then calculated.

RESULTS

Overview of estimated SSDs

A total of 10 nonionic HOCs, for which both EqP‐based and spiked‐sediment test–based SSDs could be derived, were analyzed. This included pyrethroid insecticides (i.e., bifenthrin, cyfluthrin, cypermethrin, deltamethrin, and permethrin), other insecticides (i.e., chlorpyrifos, p,p′‐dichlorodiphenyltrichloroethane [DDT], and endosulfan sulfate), and polycyclic aromatic hydrocarbons (PAHs; i.e., fluoranthene and phenanthrene; Table 1 and Figure 1). A total of 148 species were included, out of which 124 were freshwater species and 24 were saltwater species. The number of species in the EqP‐based and spiked‐sediment test–based SSDs ranged from 3 to 79 (median 13) and from 3 to 10 (median 6), respectively. The EqP‐based and spiked‐sediment test–based SSDs mostly consisted of crustacean and insect species with median proportions of 48% (range 23%–100%) and 51% (0%–77%) and 71% (67%–83%) and 25% (0%–33%), respectively. The remaining species included annelids, mollusks, ciliates, and cnidarians.

Table 1

Estimates of the 50% hazardous concentration and standard deviation values obtained from species sensitivity distributions based on the equilibrium partitioning approach and spiked‐sediment toxicity tests

			EqP‐based SSDs
				Not corrected			Period‐corrected			Spiked‐sediment test‐based SSDs
	Chemical	Log K OC a	No. of speciesb	HC50	SD	HC5	HC50	SD	HC5	No. of speciesb	HC50	SD	HC5
Pyrethroid insecticides
	Bifenthrin	6.36	4 (2, 2)	3.12	0.65	2.05	2.48	0.65	1.40	7 (5, 2)	0.61	0.76	−0.63
	Cyfluthrin	5.12	5 (3, 2)	1.95	0.65	0.88	1.35	0.65	0.28	3 (2, 1)	0.01	0.53	−0.86
	Cypermethrin	4.90	30 (8, 21)	1.18	0.80	−0.13	0.59	0.76	−0.66	5 (4, 1)	0.42	0.75	−0.82
	Deltamethrin	4.90	13 (3, 10)	1.12	0.60	0.14	0.52	0.59	−0.46	4 (3, 1)	−0.21	0.15	−0.46
	Permethrin	5.08	40 (16, 24)	2.39	0.67	1.29	1.84	0.64	0.79	6 (4, 2)	1.71	0.87	0.30
Other insecticides
	Chlorpyrifos	3.86	79 (21, 56)	1.15	0.90	−0.33	0.62	0.89	−0.85	4 (3, 1)	0.89	0.32	0.37
	p,p′‐DDT	5.23	45 (22, 23)	2.84	0.53	1.97	2.30	0.49	1.49	6 (5, 0)	2.55	0.55	1.64
	Endosulfan sulfate	3.99	3 (3, 0)	3.98	0.30	3.49	3.45	0.30	2.96	3 (2, 1)	2.15	0.79	0.85
PAHs
	Fluoranthene	4.74	13 (6, 4)	3.66	0.46	2.90	3.18	0.50	2.35	10 (7, 1)	3.53	0.50	2.71
	Phenanthrene	4.22	6 (4, 1)	3.84	0.21	3.49	3.31	0.15	3.06	6 (4, 1)	3.71	0.57	2.78

Units of HC50 and HC5 are milligrams per kilogram organic carbon after logarithmic transformation (base 10).

The K OC value was estimated by the KOCWIN program of EPI Suite, Ver 4.1.

Numbers in parentheses indicate the number of crustacean and insect species, respectively.

K OC = organic–carbon partition coefficient; EqP = equilibrium partitioning; SSD = species sensitivity distribution; HC50 and HC5 = 50% and 5% hazardous concentrations; SD = standard deviation; p,p′‐DDT = p,p′‐dichlorodiphenyltrichloroethane.

Figure 1

Species sensitivity distributions for nonionic hydrophobic organic chemicals derived from the equilibrium partitioning–based approach (green = without period correction; blue = with period correction) and spiked‐sediment toxicity tests (yellow). Lines indicate regression lines, assuming log‐normal distributions. Different symbols represent different taxonomic groups. Standard test species in spiked‐sediment toxicity tests (Hyalella azteca, Chironomus dilutus, and Chironomus riparius) are shown; 50% hazardous concentration values and their 95% confidence intervals are shown below each chemical panel. p,p′‐DDT = p,p′‐dichlorodiphenyltrichloroethane; EqP = equilibrium partitioning; oc = organic carbon.

Comparison of HC50 values derived from two approaches

Estimated HC50 values, based on the EqP approach, ranged from 13 to 9.5 × 103 mg/kg‐OC without the period correction and from 3.3 to 2.2 × 103 mg/kg‐OC with the correction. The HC50 values based on spiked‐sediment tests ranged from 0.62 to 5.1 × 103 mg/kg‐OC (Table 1). When correction for the exposure period was applied, the differences between HC50 values based on the two methods were reduced for all chemicals, except for the two PAHs (i.e., fluoranthene and phenanthrene), p,p′‐DDT, and chlorpyrifos. Moreover, the correlation coefficient between HC50 values through the two approaches increased slightly when corrected by exposure period (Pearson's r: from 0.79 to 0.81; Figure 2). Hereinafter, HC50 values based on the EqP approach without period correction were used for the two PAHs, and those with the correction were used for other chemicals to compare the HC50 values based on spiked‐sediment tests.

Figure 2

Comparison of 50% hazardous concentration (HC50) values estimated through two different approaches. Different colors represent different chemical classes. Left and right panels show equilibrium partitioning–based HC50 values without and with period correction, respectively. The solid line represents a 1:1 ratio, while broken lines represent a 1:10 and 10:1 ratio. EqP = equilibrium partitioning; p,p′‐DDT = p,p′‐dichlorodiphenyltrichloroethane; PAH = polycyclic aromatic hydrocarbon.

Differences in HC50 values between the two approaches were within a factor of 100 but reduced to within a factor of 1.7 (=100.24) when the number of species used for SSD estimation was five or more (Figure 3). In addition, the 95% CIs of HC50 values ranged across more than 2 orders of magnitude for endosulfan sulfate and cyfluthrin (Figure 1), for which spiked‐toxicity test data were available for only three species, but were reduced with an increasing number of test species. The 95% CIs of HC50 values considerably overlapped between the two approaches for the data‐rich chemicals (i.e., cypermethrin, permethrin, p,p′‐DDT, fluoranthene, and phenanthrene). Similarly, differences in HC5 values ranged from a factor of 1 to more than 129 but reduced to within a factor of 5.1 when five or more species were included (Supporting Information, Figures S3 and S4).

Figure 3

Relationships among the lower number of species in the equilibrium partitioning (EqP)–based or spiked‐sediment toxicity test–based species sensitivity distributions (SSDs) and their ratios of 50% hazardous concentration (HC50) values. The EqP‐based HC50s were derived after correction of exposure periods, except those for two polycyclic aromatic hydrocarbons without correction. Error bars indicate 95% confidence intervals calculated by considering error propagation. The dotted line represents no difference in HC50 values between the two types of SSD. The different colors represent the different chemical classes. Points are jittered along the x‐axis to avoid overlapping. PAH = polycyclic aromatic hydrocarbon.

Resampling analysis of spiked‐sediment test data demonstrated that the differences between the 97.5th and 2.5th percentile values of HC50 and HC5 values were within a factor of 3.8 (=100.58) and 45 (=101.65), respectively (Supporting Information, Figure S5).

DISCUSSION

Practical implications: EqP versus spiked‐sediment test

In the present study, our comparison of the two approaches demonstrated that when five or more species were included in each SSD, HC50 values estimated by each approach agreed suitably within a factor of 2 (not in logarithmic terms). Di Toro et al. (1991), who proposed the application of EqP theory to the derivation of sediment quality benchmarks, found that the effective concentrations (i.e., LC50 or EC50 values) in spiked‐sediment tests could be predicted to within a factor of 2–3 from K OC values and water‐only tests using the same species. This variation corresponded with the results in the present study, implying that the derivation of sediment quality benchmarks from water‐only tests based on EqP theory could be valid even when considering the differences in sensitivity among multiple species (i.e., when applied to SSD estimation). Although the EqP approach has already been used to derive sediment quality benchmarks considering SSDs in combination with the TLM (Di Toro et al., 2000), this application was limited to narcotic chemicals. In contrast, the present study provides new empirical support for the applicability of EqP theory for nonnarcotic chemicals such as pyrethroid and organophosphate insecticides while still being applicable to narcotic chemicals.

When the number of species included in the SSD was fewer than five, the differences in HC50 values based on the two approaches were more than 1–2 orders of magnitude. This result indicates that the uncertainty in sediment risk assessments with toxicity data available for fewer than five species can be considerable, whether based on SSDs using EqP theory or spiked‐sediment toxicity tests. This is also reflected by the large 95% CIs of HC50 values (e.g., cyfluthrin and endosulfan sulfate; Figure 1). Although larger uncertainties resulting from lower number of test species used in the SSD estimation have been revealed by previous studies (Hiki & Iwasaki, 2020; Newman et al., 2000), there are a limited number of standard species for spiked‐sediment toxicity tests. Therefore, if SSDs are derived based only on spiked‐sediment tests, nonstandard species must be tested. This has been exemplified by previous studies (Brock et al., 2018, 2020) and recommended by the EFSA Panel on Plant Protection Products and Their Residues (2015). In addition, standard test species in sediment tests (i.e., Hyalella azteca, Chironomus dilutus, and Chironomus riparius) were not always the most sensitive in each SSD (Figure 1), indicating that sediment risk assessments based only on such standard species can be underprotective for other sensitive species (e.g., a mayfly, Hexagenia sp., to bifenthrin [Harwood et al., 2014] and a clam, Mercenaria mercenaria, to fluoranthene [Chung et al., 2007]; Supporting Information, Table S1). Similar trends have been reported by previous SSD studies based on spiked‐sediment tests, where chronic effects in standard species such as H. azteca, C. dilutus, and C. riparius and a worm, Tubifex tubifex, were not always the most sensitive to the fungicide fludioxonil (Brock et al., 2020) and to endosulfan (Deneer et al., 2013).

Another possible explanation for the large difference observed in HC50 values between the two approaches is that effective concentrations reported in water‐only tests were not based on freely dissolved concentrations but based on total dissolved or nominal concentrations (Mokry & Hoagland, 1990; Siegfried, 1993), thus indicating that the converted was overestimated. This is likely supported by the fact that the 95% CIs of HC50 values for bifenthrin did not show overlaps between the two approaches, although this might have partly resulted from the low number of test species. Because bifenthrin is the most hydrophobic chemical examined in the present study (log K OC 6.36), it should have larger gaps between nominal (or total dissolved) and freely dissolved concentrations because of its strong binding to dissolved organic carbon in water‐only toxicity tests, leading to an overestimated HC50 value based on EqP. These results suggest that to apply EqP theory to highly hydrophobic chemicals, water‐only toxicity tests should be performed in more stable and reliable conditions, such as by using passive dosing (Fischer et al., 2016; Kwon et al., 2020), or with measurement of freely dissolved concentrations (e.g., using passive sampling [Ding et al., 2012]).

The instability and complexity of exposure concentrations in spiked‐sediment tests may be partially responsible for the difference in HC50 values between the two approaches. Concentrations of a test chemical, specifically in overlying water, in spiked‐sediment tests often fluctuate temporally because of insufficient equilibrium between the spiked sediment and surrounding water in an exposure beaker (Dorn et al., 2021; Fischer et al., 2021), insufficient equilibrium time after spiking of a test chemical into sediment (Landrum et al., 1992), or renewal of overlying water (Fischer et al., 2021; Hiki, Fischer, et al., 2021). In addition, while EqP theory assumes that porewater concentrations are representative of the toxicity to benthic organisms, benthic organisms are exposed to overlying water rather than to porewater depending on a test species and chemical (Droge et al., 2008; Whiteman et al., 1996). Because the concentration gap between overlying water and porewater can be >10 times for less hydrophobic nonionic organic chemicals (Fischer et al., 2021; Hiki, Fischer, et al., 2021), the avoidance behavior to overlying water leads to the overestimation of converted C sed,EqP. To fill the gap between water‐only and sediment tests, we recommend the detailed spatiotemporal measurement of a test chemical in spiked‐sediment toxicity tests (Fischer et al., 2021; Hiki, Fischer, et al., 2021).

The LC50 values collected from spiked‐sediment tests were highly variable, with the factors ranging from 1.4 to 130 (Supporting Information, Figure S5) even for the same chemical–species combination. Such variation led to the differences between the 97.5th and 2.5th percentile values of HC50 and HC5 values within a factor of 3.8 and 45, respectively, as demonstrated by the resampling analysis (Supporting Information, Figure S5). These differences were larger than, or comparable to, differences in HC50 and HC5 values between the EqP and spiked‐sediment test approaches. Large variation in LC50 values can be attributed to many factors including differences in testing laboratories, organism strains, and experimental conditions. Among the factors responsible for the large variation in toxicity values in spiked‐sediment tests, variation in log K OC values of the tested sediment samples and the resulting variation in bioavailability were considered to be the most influential factors (Hiki, Watanabe, & Yamamoto 2021). Therefore, our finding suggests that it is critical to consider the inherent variability observed in spiked‐sediment toxicity tests when conducting sediment risk assessments.

Exposure duration

To compare the SSDs through two different approaches, LC50 values were corrected by exposure periods (≥24 h for water‐only tests and 10–14 days for sediment tests). The period correction decreased the gap in HC50 values between the two approaches for most of the insecticides considered but increased it for PAHs. This is reasonable because the effective concentrations of narcotic chemicals, including PAHs, reach a constant value after the internal body concentration reaches an equilibrium with the external concentration (Escher & Hermens, 2002; Legierse et al., 1999). Ninety‐six hours are enough to reach equilibrium for aquatic organisms (Di Toro et al., 2000). In contrast, the effective concentrations of chemicals exhibiting an irreversible receptor interaction, including insecticides, markedly decrease with increasing exposure periods (Escher & Hermens, 2002; Legierse et al., 1999). As shown in the present study, period correction based on the correlation between the exposure period and toxicity was effective in comparing HC50 values between two approaches; however, a highly mechanistic correction, such as using a toxicokinetic–toxicodynamic model (Jager et al., 2011; Kon Kam King et al., 2015), is also possible if more abundant ecotoxicity data are available.

Taxonomic composition

In the present study, the EqP‐based HC50 values were larger than the spiked‐sediment test‐based HC50 values by a factor >10 for all chemicals with fewer than five test species (i.e., bifenthrin, endosulfan sulfate, and cyfluthrin). This is likely because crustaceans are generally more sensitive to pyrethroid, organochlorine, and organophosphate insecticides than insects (Giddings et al., 2019; Wang et al., 2019), and the spiked‐sediment tests analyzed were biased toward crustacean species (Table 1). In contrast, crustaceans, in particular cladocerans, are generally less sensitive than insects to neonicotinoid insecticides (Raby et al., 2018), which were not analyzed in the present study. These sensitivity differences among taxonomic groups indicate the need to include test species belonging to various taxonomic groups in SSD estimation (e.g., Brock et al., 2018).

When a clear gap exists in the chemical sensitivity among different taxonomic groups, the use of a simple unimodal model may be problematic because of poor fitting in the left tail regions of the SSD (Fox et al., 2021). To address this, several methods have been proposed to estimate hazardous concentrations, such as using only the data from the most sensitive taxonomic groups (e.g., insects) or using a multimodal model (Del Signore et al., 2016; Fox et al., 2021; Nagai, 2016). Although the present study and others on spiked‐sediment tests (Brock et al., 2018, 2020) have used a single log‐normal distribution SSD largely because of the limited availability of toxicity data, the proposed methods may be more beneficial in analyzing and comparing SSDs based on EqP theory and spiked‐sediment toxicity tests.

Comparison with other sediment quality benchmarks

Because many other sediment quality benchmarks have been proposed for PAHs (Long et al., 1995; Macdonald et al., 1996; McGrath et al., 2019; USEPA, 2003; Verbruggen, 2012) and insecticides (Nowell et al., 2016), such values were compared with the HC5 values estimated in the present study (Figure 4).

Figure 4

Comparison of sediment quality benchmark values derived from equilibrium partitioning (EqP) theory and/or spiked‐sediment toxicity tests. Threshold effect benchmark (TEB) and likely effect benchmark (LEB) were taken from Nowell et al. (2016). Effect range low (ERL) and threshold effect level (TEL) were taken from Long et al. (1995) and Macdonald et al. (1996) and converted from milligrams per kilogram to milligrams per kilogram organic carbon, assuming 1% organic carbon content. Maximum permissible concentrations (MPC) were taken from Verbruggen (2012) and converted from milligrams per kilogram to milligrams per kilogram organic carbon, assuming 5.8% organic carbon content (i.e., 10% organic matter). The MPC values for fluoranthene and phenanthrene were derived from chronic sediment toxicity data. The EqP sediment benchmark was taken from US Environmental Protection Agency (2003). Only period‐corrected 5% hazardous concentration (HC5) values based on EqP theory are shown for insecticides, while only HC5 values without correction are shown for polycyclic aromatic hydrocarbons (i.e., fluoranthene and phenanthrene). p,p′‐DDT = p,p′‐dichlorodiphenyltrichloroethane; ESB = EqP sediment benchmark.

Effect range low (ERL) and threshold effect level (TEL) are estimates of threshold contamination levels derived from an analysis of field‐collected contaminated sediment and correspond to the 10th and 15th percentiles of effects data, respectively (Long et al., 1995; Macdonald et al., 1996). These are based on dry weight concentrations (i.e., milligrams per kilogram) and thus were converted to an organic carbon weight basis (i.e., milligrams per kilogram organic carbon), assuming 1% organic carbon for the comparison with our HC5 values. For both fluoranthene and phenanthrene, HC5 values in the present study were higher than ERL and TEL values by a factor >5, regardless of whether they were based on EqP theory or spiked‐sediment toxicity tests (Figure 4). This is reasonable, given that these empirical benchmarks (i.e., ERL and TEL) are derived from toxicity tests and observations using field‐collected sediments, which contain many contaminants. The overprotection of empirical benchmarks has been pointed out by numerous previous studies (Di Toro et al., 2000; McGrath et al., 2019).

Maximum permissible concentrations (MPC), threshold effect benchmark (TEB), and likely effect benchmark (LEB) were derived from spiked‐sediment toxicity tests, where the test chemical was the only causative agent (Nowell et al., 2016; Verbruggen, 2012). Whereas MPC and TEB intend to represent effective concentrations in chronic toxicity tests, LEB represents those in subchronic toxicity tests. Generally, MPC and TEB values were lower than the HC5 values derived in the present study, and LEB values were higher than the HC5 values except for chemicals with a lower number of test species (e.g., EqP‐based HC5 for bifenthrin). The lower MPC values may be due to application of the AF of 10 to the results of chronic tests with annelids, crustaceans, and insects. The fact that most of the HC5 values in the present study fell between TEB and LEB values is reasonable, considering that TEB and LEB were derived from chronic and subchronic spiked‐sediment tests with standard test species (i.e., H. azteca and Chironomus spp.), respectively.

For PAHs, EqP sediment benchmarks (ESBs) were derived from water‐only test data using the TLM, EqP theory, and SSD (USEPA, 2003). The ESB values were comparable to the HC5 values derived in the present study, based on spiked‐sediment toxicity tests by a factor of <1.5 (Figure 4). While ESB values were derived from acute water‐only tests but were converted to chronic estimates using the acute‐to‐chronic ratio, HC5 values in the present study were based on lethality tests within 10–14 days (i.e., subchronic tests). Agreement between our HC5 and ESB implies that subchronic sediment tests can be representative for chronic effects of at least narcotic toxicity, which is independent of time after reaching equilibrium as described in the Exposure duration section. In addition to narcotic chemicals, the difference between the LC50 values in subchronic and the no‐observed‐effect concentration values in chronic sediment tests (with periods of >28 days) has been reported to be within a factor of 5 for several chemicals (Anderson et al., 2015; Deneer et al., 2013) and field‐collected sediments (Ingersoll et al., 2015). In contrast to HC5 values based on spiked‐sediment toxicity tests, the HC5 value based on EqP theory without period correction was 5 times higher than ESB value for phenanthrene, although it was comparable to the ESB value for fluoranthene.

The ESB value has proven to be protective of the lethality in H. azteca exposed to field‐collected sediment samples, with a false‐negative rate of 0.5% (McGrath et al., 2019) and protective of the abundance of a marine amphipod, Ampelisca abdita, in a PAH‐contaminated field (Di Toro et al., 2000). Thus, agreement between ESB and the HC5 value derived in the present study provides modest evidence that SSDs based on spiked‐sediment toxicity tests can provide threshold concentrations below which minimal effects on benthic communities are expected. Because HC5 values were consistent between EqP theory and spiked‐sediment tests when the number of species was sufficient (Supporting Information, Figure S4), SSDs derived based on EqP theory, if necessary with period correction (see above, Exposure duration), could be used to estimate protective sediment benchmarks. Further research is needed to examine this expectation for a variety of chemicals based on field surveys on benthic community and toxicity tests using field‐collected sediment samples.

The ESB has also been demonstrated to be overly protective, such that field‐collected sediment samples which exceeded ESB values for PAHs are often not toxic (McGrath et al., 2019). This is due to the fact that black carbons in sediment have strong binding affinity to PAHs and reduce the bioavailability of PAHs more than what is expected from sediment concentrations (Endo et al., 2020; McGrath et al., 2019). Benchmarks based on sediment concentrations, including the estimation in the present study, can thus be used as screening‐level risk assessments but may lead to an overestimation of ecological risks. In a higher‐tier assessment, more accurate methods, such as passive sampling techniques, should be used to directly quantify the bioavailability (Lydy et al., 2014; McGrath et al., 2019), thereby refining sediment risk assessments.

Limitations and further investigations

The present study compared SSDs for nonionic HOCs using two different approaches: EqP theory and spiked‐sediment toxicity tests. It was demonstrated that when the number of species was five or more, the difference in HC50 values between the two approaches was within a factor of 2. This suggests that the EqP approach, which usually has toxicity data for a large number of species, or the combined use of the two approaches can be used to derive sediment quality benchmarks in combination with SSD estimation as well as spiked‐sediment toxicity tests alone. However, because only 10 HOCs with 10–14‐day sediment LC50 values were analyzed in the present study, further studies with additional spiked‐sediment toxicity data are necessary, particularly with chronic toxicity data and data for noninvertebrate species. To account for the diversity of species sensitivity in benthic field environments, spiked‐sediment toxicity tests should be applied to a wide range of species, including benthic fish (Schreiber et al., 2018), and not solely invertebrate species for which the current tests are biased. This will undoubtedly require further development and standardization of test methods for various benthic organisms.

Supporting Information

The Supporting information is available on the Wiley Online Library at https://doi.org/10.1002/etc.5270.

Disclaimer

The authors declare that there are no conflicts of interest. The opinions expressed in this article do not necessarily represent the official views of the authors' affiliations.

Author Contributions Statement

Kyoshiro Hiki designed the study, curated the toxicity data, and wrote the draft of the manuscript. Kyoshiro Hiki and Yuichi Iwasaki calculated and evaluated the data. Yuichi Iwasaki, Haruna Watanabe, and Hiroshi Yamamoto reviewed and revised the manuscript.

This article has earned an Open Data & Open Materials badge for making publicly available the digitally shareable data necessary to reproduce the reported results. The data are available at https://github.com/KyoHiki/EqP_vs_Spiked_sediment_test. Learn more about the Open Practices badges from the Center for Open Science: https://osf.io/tvyxz/wiki.

This article includes online‐only Supporting Information.

Supporting information.

Click here for additional data file.

Supporting information.

Click here for additional data file.