Published on the Web 7/29/2009.
Probabilistic ecological hazard assessment of parabens using Daphnia magna and Pimephales promelas†
Article first published online: 6 JAN 2010
Copyright © 2009 SETAC
Environmental Toxicology and Chemistry
Volume 28, Issue 12, pages 2744–2753, December 2009
How to Cite
Dobbins, L. L., Usenko, S., Brain, R. A. and Brooks, B. W. (2009), Probabilistic ecological hazard assessment of parabens using Daphnia magna and Pimephales promelas. Environmental Toxicology and Chemistry, 28: 2744–2753. doi: 10.1897/08-523.1
- Issue published online: 6 JAN 2010
- Article first published online: 6 JAN 2010
- Manuscript Accepted: 25 MAR 2009
- Manuscript Received: 18 OCT 2008
- Ecological risk assessment;
- Comparative toxicity;
- Chemical toxicity distribution;
- Personal care product
Parabens are common antimicrobial agents found in thousands of pharmaceuticals and personal care products. Parabens are introduced into aquatic ecosystems from wastewater treatment plant effluents and have been detected in surface waters in the low microgram per liter range. Although these compounds display low toxicity in mammals, paraben toxicity to aquatic organisms has not been investigated. Standardized acute and subchronic endpoints in larval fish (Pimephales promelas) and cladoceran (Daphnia magna) models were examined for seven different parabens (methyl-, ethyl-, isopropyl-, propyl-, isobutyl-, butyl-, benzylparaben), which encompassed a range of log P values. Paraben 48 h median lethal concentration values (LC50) ranged from 4.0 to 24.6 mg/L in D. magna and 3.3 to > 160.0 mg/L in fathead minnow. Growth and reproduction in D. magna had lowest-observed-effect concentrations (LOECs) ranging from 0.12 to 9.0 mg/L and 1.5 to 6.0 mg/L, respectively. Fathead minnow growth was adversely affected at levels ranging from 1.0 to 25.0 mg/L. Aquatic toxicity of the parabens was inversely related to lipophilicity, suggesting that responses using standardized endpoints resulted from narcosis. Utilizing toxicity benchmark concentrations (e.g., LC50s, LOECs) for each compound, chemical toxicity distributions, a probabilistic hazard assessment technique, were developed to assess the probabilities of detecting parabens that elicit a response at or below a given concentration. For the responses assessed in the present study, the 5th centile values (the concentration at which 5% of parabens elicit a response) ranged from 15 μL to 2.43 mg/L, with D. magna growth eliciting the lowest 5th centile value and acute D. magna mortality eliciting the highest. The distributions demonstrated that at environmentally relevant concentrations in developed countries there is limited acute or subchronic aquatic hazard of parabens to the organisms and responses examined.
Parabens are alkyl hydroxyl benzoate preservatives used to prevent microbial growth in many common care products, such as cosmetics, underarm deodorants, hair products, pharmaceuticals, and some foods. A 1984 report by Elder  estimated paraben use in over 13,200 cosmetic products. The European directive for cosmetics limit concentrations of individual parabens to 0.4% of the cosmetic formulation, but allow mixtures of parabens up to 0.8% of the formulation . The U.S. Food and Drug Administration regulations limit paraben concentrations in foods to 0.1% [3,4], while paraben concentrations in pharmaceuticals rarely exceed 1% .
Production estimates for parabens have been based on disappearance data, the amount distributed at a national level, reported by producers. For 1987, the estimated usage in the United States was 6,947 kg and 424 kg for propylparaben and methylparaben, respectively. It is estimated that <7,727 kg of total parabens are produced each year in the United States . Because parabens are widely used in many food, cosmetic and pharmaceutical formulations, they have recently been detected in wastewater effluent, surface waters, and even human breast cancer tumors [7−9]. It appears that introduction of parabens to aquatic ecosystems primarily occurs from decentralized wastewater treatment systems or centralized municipal wastewater treatment plants (WWTPs). The WWTPs have been shown to be effective at removing the majority of parabens in influent water in developed countries , but parabens have been detected in environmental samples up to low microgram per liter levels. A study in 2005 examined concentrations of methyl-, ethyl-, propyl-, and butylparaben in influent and effluent samples from eight WWTPs in southern Ontario, Canada. Paraben concentrations in influent samples ranged from 0.02 to 2.43 μL while effluent concentrations ranged from <0.01 to 0.05 μL . Another study analyzed methyl-, ethyl-, and propylparaben in river water, WWTP effluent, and industrial effluent: river water contained the highest concentrations of parabens (85.1, 53.8, and 78.4 ng/L for methyl-, ethyl-, and propylparaben, respectively), while WWTP and industrial effluent concentrations ranged from 2.1 to 8.5 ng/L . More recently, a study found maximum concentrations of methylparaben, ethylparaben, propylparaben, and butylparaben to be 400, 15, 24, and 52 μL, respectively, in surface waters in South Wales, United Kingdom . Environmental levels of parabens have not been explored in less developed countries with limited WWTP infrastructure.
The U.S. Food and Drug Administration generally recognizes methyl- and propylparabens as safe for humans due to their low mammalian toxicity . However, limited information is available for the potential effects of parabens to aquatic organisms. One study reports median effective concentration (EC50) values of 11.2 and > 50 mg/L for Daphnia and fish, respectively, exposed to methylparaben . Beyond this report, there are limited data available for aquatic organisms. A metabolite of parabens, p-hydroxyben-zoic acid, was used to examine potential growth effects on a freshwater green alga Pseudokirchneriella subcapitata. Algal growth was stimulated at the lower concentrations of 13.8 and 138 mg/L, but growth inhibition was seen at higher concentrations, with a median growth inhibition concentration (IC50) of 1.37 g/L . Eklund  observed growth inhibition in four strains of bacteria (Escherichia coli, Pseudomonas American Type Cell Collection [ATCC] 9027, Baccillus subtilis ATCC 6633, and Baccillus subtilis W23) after exposure to methyl-, ethyl-, propyl-, and butylparaben. Butylparaben was the most effective at inhibiting bacteria growth, with IC50 values ranging from 29.1 to 194 mg/L in the different bacterial strains. Methylparaben was the least effective at inhibiting growth, with IC50 values ranging from 548 to 837 mg/L . A similar study exposing Klebsiella pneumoniae and Saccha-romyces cerevisiae to methyl-, ethyl-, propyl-, butyl-, pentyl-, and hexylparben showed similar results. Again, methylparaben was the least toxic to the bacteria (IC50 values of 680 and 1,042 mg/L), whereas hexylparaben was the most toxic paraben (IC50 values of 14 and 44 mg/L) .
Parabens have been demonstrated to elicit estrogen agonist responses after activating the estrogen receptor both in vivo and in vitro. In vitro, by use of the yeast estrogen screen, methyl-, ethyl-, propyl-, and butylparaben were shown to have estrogenic properties at least 10,000-fold less than 17β-estradiol . The MCF-7 (human ademocarcinoma cell line) assay also showed proliferation after exposure to methyl-, ethyl-, isopropyl-, propyl-, isobutyl-, butyl-, and benzylparaben at concentrations 105 to 107 higher than 17β-estradiol [16–19]. In vivo, rat uterotrophic assays showed that isobutyl-, butyl-, and benzylparaben all elicited estrogenic responses after subcutaneous administration [15,17,18]. Propylparaben has also demonstrated vitellogenin induction in Japanese medaka (Oryzias latipes) after aqueous exposure at 9.9 mg/L and higher .
Because standard acute and chronic toxicological data are limited to mostly mammalian and microorganism models, it is currently difficult to characterize paraben hazard to aquatic life. Therefore, the goal of the present study was to perform a hazard assessment for parabens using common invertebrate and vertebrate models and responses. Specifically, the objectives of the present study were to define acute and subchronic toxicity thresholds for seven parabens using common vertebrate and invertebrate models; to examine whether there is a relationship between aquatic toxicity and lipophilicity of parabens; and to use a probabilistic chemical toxicity distribution (CTD) approach to characterize hazards associated with parabens in aquatic environments. We hypothesized that toxicological benchmark concentrations for these model organism responses would decrease with increasing paraben lipophilicity (as log P).
MATERIALS AND METHODS
Parabens used in the present study included methylparaben, ethylparaben, isopropylparaben, propylparaben, isobutylpar-aben, butylparaben, and benzylparaben. All compounds were white solids with purities ⩾99%. Methylparaben, ethylparaben, propylparaben, butylparaben, and benzylparaben were purchased from Sigma Aldrich. Isobutylparaben was purchased from Wako Chemicals, and isopropylparaben was obtained from AK Scientific. Structural formulas for the parabens used are presented in Figure 1. The lipophilicity (log P) of each paraben examined in the present study generally increased with increasing chain length. Branching of the chain results in a lower lipophilicity than the same molecule without chain branching. Physical and chemical properties of the seven study parabens are summarized in Table 1.
|Paraben||CAS no.a||Molecular wt (g/mol)||log P||Solubility (mg/L)|
Acute test methods
Acute toxicity tests were performed in triplicate using 24 h posthatch Pimephales promelas and <24 h old Daphnia magna neonates following methods previously described . Static, 48 h exposure to five concentrations of each paraben were conducted with each species in a temperature-controlled incubator at 24°C with a 16:8 light:dark cycle. Reconstituted hard water (RHW) controls and 0.1% acetone solvent controls were included in each test. All RHW used for testing had the following range of water quality parameters: pH = 8.2 to 8.8; dissolved oxygen = 7.1 to 9.0 mg/L; conductivity = 535 to 588 μS/cm; alkalinity = 110 to 120 mg/L as CaCO3; and hardness = 160 to 175 mg/L as CaCO3. To make test solutions, parabens were dissolved in high-performance liquid chromatography grade acetone and delivered into RHW at 0.1% acetone. For each concentration, two replicates of 10 P. promelas in 250 ml of test solution were exposed in 500 ml beakers without food for 48 h. Four replicates of five D. magna in 80 ml of test solution were exposed without food in 100 ml beakers. No aeration was used in the test chambers, per U.S. Environmental Protection Agency (U.S. EPA) recommendations. Mortality was documented at 24 and 48 h. The LC50 was calculated for each exposure.
Subchronic test methods
Following the U.S. EPA methods, subchronic toxicity tests were conducted using 24 h posthatch P. promelas larvae  and ≤24 h old D. magna neonates . Half of the LC50 determined from the acute test was used as the highest subchronic exposure concentration; a 0.5 dilution factor was then applied to prepare the remaining concentrations. High-performance liquid chromatography grade acetone was used to deliver the parabens to the test solutions at °0.1% acetone. All RHW used for testing had the following water quality parameters: pH, 8.2 to 8.8; dissolved oxygen, 7.1 to 9.0 mg/L, specific conductance = 535 to 588 μS/cm, hardness = 160 to 175 mg/L as CaCO3, and alkalinity = 110 to 120 mg/L as CaCO3. All tests were maintained in a temperature-controlled incubator at 25°C with a light cycle of 16:8 h light:dark.
Seven day, static-renewal exposures to five concentrations (plus a RHW and 0.1% acetone solvent control) of each paraben were conducted using P. promelas. Four replicates of 10 P. promelas in 250 ml of test solution for each concentration were prepared and renewed daily. The fish were fed newly hatched Artemia nauplii twice daily. Dissolved oxygen was measured in the outwater to assure that oxygen levels were not falling to intolerable levels. During the exposures, dissolved oxygen levels never decreased below 5 mg/L. Mortalities were observed on a daily basis. After 7 d of exposure, the remaining live organisms in each replicate were pooled together, dried, and weighed on a Mettler-Toledo Model MX5 microbalance after desiccation was complete. The average weight per fish was calculated and compared to the control.
Ten day, static-renewal exposures were conducted in a similar fashion with D. magna following U.S. EPA methods  with minor modifications previously reported [24–26]. Ten replicates with one Daphnia in 80 ml of each concentration of test solution (plus a RHW and 0.1% acetone solvent control) were prepared and renewed every other day. The daphnids were fed a mixture of Cerophyll® (ScholAR® Chemistry) and P. subcapitata daily [27,28]. Mortalities and number of neonates produced were recorded daily. At the end of each 10 d study, surviving mothers were desiccated and weighed on a Mettler-Toledo Model MX5 microbalance  for comparison to the controls. Total neonates produced per daphnid were calculated and compared to controls.
For all acute tests, LC50 values were calculated for each paraben using the probit method, if assumptions were met, or a trimmed Spearman-Karber method, if assumptions for the probit method were not met. Because all acute tests were run in triplicate, the average LC50 value from the three exposures was calculated. For each compound used for chronic exposures, the lowest-observed-effect concentration (LOEC) was calculated. A t test was used to compare the RHW control to the solvent control. If there was no significant difference (p < 0.05), the two controls were pooled. To derive the LOEC value, the data for each paraben were first analyzed using Shapiro-Wilk's test for normality. If the data were normally distributed, Bartlett's test for homogeneity of variance was conducted. For data with homogenous variances, Dunnett's test was run when the number of replicates was equal, and a t test with a Bonferroni adjustment was run if replicates were unequal (such as when an organism did not survive the duration of the exposure). When the data were non-normally distributed and/or had heterogeneous variance, Steel's many-one rank test was used for tests with equal replication, while the Wilcoxon rank sum test with Bonferroni adjustment was used when there was unequal replication. Statistically significant differences from controls were determined at α = 0.05. Average statistical power (β = 0.2; α = 0.1) for treatment-control comparisons was calculated for each paraben for D. magna growth and reproduction, and fathead minnow growth tests using Sigmastat® 2.0 (SPSS Jandel).
Analytical verification of aqueous concentration
To verify aqueous concentration of parabens for chronic tests, water samples of the freshly prepared test solution for each concentration were extracted onto solid phase extraction (SPE) columns on day zero and the last day of renewals. Additionally, outwater of the first renewal (day 1 for fathead minnow tests and day 2 for D. magna tests) from all replicates was pooled for extraction. Sample preparation and analysis followed protocols described by Canosa et al. . Water samples were spiked with isotopically labeled 13C6p-hydroxybenzoic acid surrogate (Cambridge Isotope Laboratories) and then adjusted to pH 2.5 with nitric acid. The test water was extracted onto Waters Oasis® Hydrophilic Lipophilic Balanced SPEs (6 cc, 20 mg) at a rate of 5 ml/min. Before extraction, the columns were conditioned with 5 ml of ethyl acetate, followed by 5 ml of methanol, then 5 ml of nanopure water adjusted to pH 2.5 . After extraction, SPEs were stored at −20°C until analysis.
To elute the samples off of the SPEs, the columns were first thawed at room temperature. At a rate of 5 ml/min, 4 ml of ethyl acetate were drawn through the SPEs and collected. Analytes were derivatized by adding 40 μl of N-methyl-N-(tert-butyldimethylsilyl)-trifluoroacetamide to the collected ethyl acetate . The ethyl acetate solution was dried with sodium sulfate and stored at 4°C until analysis. Approximately 50% of the collected samples were analyzed using gas chromatography (GC)/mass spectrometry (MS)/MS, including highest, middle, and lowest treatment levels.
Concentrations of parabens and their degradation product (p-hydroxybenzoic acid) were determined using a Varian CP-3900 GC coupled with a Varian Saturn 2100T MS/MS (ion trap) equipped with a Varian CP-8400 autosampler (Varian). Analytes were chromatographically separated using a VF-5 MS (30 m × 0.25 mm × 0.25 m) capillary column. The Saturn mass spectrometer was operated at 200°C trap temperature, and a manifold temperature of 110°C. The column temperature gradient was as follows: initial temperature, 100°C; ramped to 180°C at 15°C/min and held for 5 min; and ramped to 290°C at 6°C/min and held for 31 min. Helium was used as carrier gas at a constant flow rate of 1 ml/min (linear velocity 37.2 cm/s). Injections of 1.0 μl were made using splitless mode and an injection port temperature of 275°C. The transfer line was kept at 280°C. Paraben concentrations were calculated using mirex (99.9% pure; Supelco) as the internal standard.
Traditional hazard quotients (HQs) were calculated for each paraben to determine the magnitude of difference between the exposure value and the effects value. For each paraben, the exposure values from the literature [7,8,10] were averaged to get a measured environmental concentration. For parabens that did not have literature exposure values, the average of all exposure values for all parabens was derived and used as the predicted environmental concentration. To calculate the HQ, the measured environmental concentration or predicted environmental concentration was divided by the most sensitive no-observed-effects concentration (NOEC) for each paraben. When a hazard quotient is greater than one, there is potential for toxic effects in the environment, but when the ratio is less than one, toxicity is unlikely.
Additionally, after the completion of each acute and chronic toxicity assay, an LC50 or LOEC value was calculated. For each assay, the corresponding LC50 or LOEC values for the seven parabens were fitted to log-normal distributions according to Burmaster and Hull  described in Solomon et al. , Dobbins et al. , and Brain et al. . This was accomplished by ranking the data in increasing order and calculating a percent rank for each concentration using a Weibull formula:
where j is the percent rank, i is the rank assigned to the LC50 or LOEC value after placing the values in increasing order (1 to n), and n is the number of compounds. The concentrations were plotted on a logarithmic scale against their respective percent rank, on a probability scale. A regression line was added to develop a line of best fit for the percent rank and concentration data.
Probability calculations were made by modifying the equations from Solomon et al.  by substituting a single threshold value instead of percentage-based exposure values using Microsoft Excel® 2003 as
Probability below x
where NORMSDIST returns the standard normal cumulative distribution function, x is the threshold value, and mtox and btox are the slope and intercept, respectively, of the regression line of the CTD . Based on this technique, 5th centile values were calculated for each assay, and the probability of encountering a paraben that elicits effects at or below 0.1 μL was also calculated. Paraben concentrations in the environment range from 0 to 0.4 μL, based on the studies cited within this paper. The trigger value of 0.1 μL falls within this range, and using an environmental concentration on the high end of the range results in a more conservative effect probability.
All calculations of LC50 values were based on nominal concentrations. Control mortality in RHW and solvent never exceeded 10%. Additionally, there was no significant difference (p > 0.05) between the RHW control and solvent control in any test. In both species, significant relationships (p < 0.01) between log P and LC50 values for the parabens were demonstrated (Fig. 2A). During the acute studies, organisms were first observed either immobilized or sluggish, and with time, the parabens ultimately led to the death of the organism.
The LC50 values for D. magna and fathead minnow ranged from 4.0 to 24.6 mg/L and 3.0 to >160 mg/L (Table 2), respectively, with methylparaben being the least toxic and benzylparaben being the most acutely toxic to both model organisms. The highest concentration of methylparaben, 160 mg/L, elicited partial mortality (between 10 and 75% mortality, mean mortality 40%, n = 3) in the fathead minnow model after 48 h of exposure, but did not elicit notable mortality at lower concentrations (e.g., 10–120 mg/L). The 160 mg/L concentration was the highest treatment level that would visually remain in solution (methylparaben has an estimated aqueous solubility of 4.2 g/L [Table 1], but in our studies, this level could not be reached with the compound remaining in solution); thus, we could not estimate an LC50 value for fathead minnows exposed to methylparaben.
The results from the acute toxicity tests show similarities in sensitivities to parabens between the D. magna and the fathead minnow models (Table 2). Differences in sensitivities between study organisms exist for methylparaben, where the fathead minnow is distinctly less sensitive than D. magna. Methylparaben did not affect the fathead minnow enough to cause significant mortalities up to 160 mg/L. However, 24.6 mg/L was the derived LC50 value for D. magna after exposure to methylparaben. On the other hand, compounds with the highest log P values (isobutylparaben, butylparaben, and benzylparaben) elicited similar LC50 values in both D. magna and fathead minnow (Table 2).
The highest methylparaben treatment level in the D. magna chronic test (12 mg/L) was the only treatment level to significantly increase mortality (60% mortality observed) compared to the controls. For all chronic tests, mortality in RHW control and solvent controls did not exceed 10%, and the two control treatments never differed significantly (p > 0.05). Average statistical power for the chronic tests ranged from 0.43 to 1.0 (Table 2).
Daphnia magna growth LOEC values ranged from 0.1 to 9 mg/L, while fathead minnow LOECs ranged from 1.0 to 25 mg/L (Table 2). Additionally, growth LOEC values for both species were related (p < 0.01) to log P of the parabens (Fig. 2B). Daphnia magna reproduction was also used as an endpoint of effects after paraben exposure, but there was no statistically significant relationship (p = 0.59) between the LOEC for reproduction effects and paraben log P values. There was little variation in the different paraben LOEC values for D. magna reproduction (Table 2), but this was the most sensitive endpoint for the least lipophilic parabens (methylparaben, ethylparaben, and isopropylparaben).
Mean measured paraben concentrations in treatment levels generally showed good agreement with nominal concentration in all tests performed. Approximately 40 to 60% of the water treatments were measured for each test exposure. Measured paraben concentrations ranged from 60 to 115% of nominal for the dilution water, for both D. magna and fathead tests. While the freshly prepared test solutions varied very little from nominal concentrations, the outwater from D. magna tests (48 h renewal) ranged from 24 to 72% of nominal concentrations, and outwater from fathead minnow tests (24 h renewal) ranged from 70 to 96% of nominal. Parabens measured in blank samples were less than 0.1% (∼1 ng/L) of paraben mass measured in treatment solutions. The limit of quantification for all paraben analytes were 0.005 ng/μl. Average surrogate recoveries in the blanks were 93% with a relative standard deviation of 6.84%.
The majority of samples were verified analytically to be 90 to 100% of nominal treatment levels. There were a few exceptions, such as with the highest concentration of methylparaben in the D. magna chronic test, which was a lower percentage of nominal (60%) than the lower methylparaben concentrations. The lower methylparaben concentrations were diluted from the highest methylparaben concentration, but were 90 to 100% of nominal. This deviation from nominal of the highest methylparaben treatment was shown in both day zero and final day dilution water. This could potentially be due to overloading of the SPE column because the amount of methylparaben going onto the column was higher than other paraben concentrations used in the exposures. This test employed extracting 1 L of test water, while most other tests only extracted 100 ml of test water. Therefore, the amount of paraben accumulating on the SPE was greater than other paraben tests. Additionally, there was a significant drop in percentage of nominal in the outwater samples taken from the test during renewal. These concentrations were much lower than nominal, which could be attributed to photolysis, sorption to the test vessel or uptake by the organisms. Because the outwater concentrations were lower in the D. magna tests (24–72% of nominal vs 70–96% of nominal in fathead minnow tests), the loss of parabens in the test solutions is most likely due to photolysis. Daphnia magna test solutions were renewed every other day, meaning that the outwater had been exposed to light for 2 d. Conversely, the fathead test solutions were renewed daily; thus, outwater from these toxicity tests were only exposed to light for 24 h.
|Chronic LOEC values (mg/L) and average statistical power|
|Mean (n = 3) acute LC50 values and (SD) (mg/L)||D. magna growth||D. magna reproduction||Fathead growth|
|D. magna||Fathead||LOEC||Power||LOEC||Power||LOEC||Power||Hazard quotient|
|Methylparaben||24.6 (8.3)||>160b||6.0||0.993||1.5||0.949||25.0||0.813||9.0 × 10-5|
|Ethylparaben||18.7 (5.7)||34.3 (1.6)||9.0||0.480||2.3||0.450||17.0||0.769||7.8 × 10-6|
|Isopropylparaben||8.5 (0.4)||17.5 (1.3)||4.0||0.837||2.0||0.999||9.0||0.548||2.8 × 10-5|
|Propylparaben||12.3 (4.3)||9.7 (0.6)||0.4||0.654||6.0||0.996||2.5||0.442||3.1 × 10-5|
|Isobutylparaben||7.6 (4.7)||6.9 (1.7)||0.3||0.483||2.0||0.596||3.5||0.667||1.1 × 10-4|
|Butylparaben||5.3 (0.9)||4.2 (1.5)||0.2||0.587||2.6||1||1.0||0.587||6.5 × 10-5|
|Benzylparaben||4.0 (1.8)||3.3 (1.4)||0.1||0.677||2.0||0.998||1.7||0.430||2.3 × 10-4|
In deriving HQs for each paraben, the NOEC value from the most sensitive assay for each paraben was used to give the most protective HQ. Even using the most sensitive NOEC, HQs ranged from 7.8 × 10-6 to 2.3 × 10-4 (Table 2). Because these ratios are much smaller than 1, this is highly indicative of limited toxic effects of parabens to fathead minnow and D. magna in the environment.
Effect distributions for fathead and D.magna assays are presented in Figure 3A to E, and the data used to generate the distributions are summarized in Table 2. Utilizing the CTD linear regression equations, threshold values and the probability of finding a compound in the environment that elicits a given response at or below 0.1 μL were derived using probabilistic analysis. For each assay, 5th centile values were calculated to estimate the concentration at which five percent of all parabens elicit the given response. The 5th centile values were 0.015, 0.37, 0.74, 0.98, and 2.43 mg/L for D. magna growth, fathead growth, fathead mortality, D. magna reproduction, and D. magna mortality, respectively. When determining the probabilities of finding parabens that will elicit detrimental effects on growth and reproduction of D. magna and fathead minnow at or below an environmentally realistic concentration of 0.1 μL, all CTDs predicted less than 0.1% chance.
The primary objective of the present study was to test the hypothesis that paraben toxicity to a model cladoceran and fish was related to log P values of parabens. We observed significant relationships between fish and cladoceran mortality and growth and log P (Fig. 2). A possible mode of action of toxicity for parabens on the aquatic organisms tested is narcosis. Narcosis, often referred to as baseline aquatic toxicity, is characterized by an initial decrease in organism activity and response to stimuli, and can often lead to death . This decrease in activity was observed in organisms in paraben acute toxicity tests. The Meyer-Overton rule states that the narcotic effect will increase with increasing lipophi-licity (measured as log P or by similar partitioning coefficients) . The relationship between paraben toxicity and lipophi-licity demonstrated what is expected under the Meyer-Overton rule, and thus lends support to narcosis as a potential mode of action for parabens effects to these organisms and endpoints.
Other studies have used the ratio of the 24 h LC50 value to the end of the test period LC50 (e.g., 24-h LC50:96-h LC50 ratio) to examine the mode/mechanism of action in aquatic toxicity tests. In this context, a ratio of near 1 has been attributed to narcosis; ratios of greater than 2 were attributed to a different mode of action . For the present paraben study, in the fathead minnow, the average 24-h LC50:48-h LC50 ratios for each paraben were all less than 2, indicative of narcosis. However, in the D. magna acute mortality model, the less lipophilic compounds (methyl-, ethyl-, and isopropylparaben) all had ratios greater than 2, potentially indicating a different mode of toxic action in invertebrates. This observation supports the lack of relationship observed between LOEC values for D. magna reproduction and log P values.
Both fathead minnow and D. magna growth were affected by parabens in the mg/L range, but D. magna growth was more sensitive to paraben exposure than the fathead minnow growth responses. Daphnia magna responded to parabens as low as 0.1 mg/L whereas fathead minnows only responded at 1 mg/L and higher. Consequently, D. magna appear to be more susceptible to effects from parabens in the environment. Additionally, parabens that are more lipophilic appear more likely to elicit responses in the environment than those that are less lipophilic. Methylparaben and propylparaben are the most frequently used parabens in cosmetic formulations. Both compounds are often used in combination to maximize the antimicrobial effect because propylparaben appears to have a higher level of antimicrobial effects, though methylparaben is the most soluble . Because these two parabens are relatively less lipophilic, and thus not as toxic as some of the more complex parabens, the two most widely used parabens may present lower ecological risk than benzylparaben, which is not used as frequently .
As chain length of parabens increase, lipophilicity and toxicity to bacteria may be expected to concomitantly increase. A study in 1979 synthesized straight chain parabens up to 18 carbons long. The minimum growth inhibition concentration decreased 3,000-fold as chain length increased, up to the 12 carbon chain paraben . Though longest chained parabens are not commercially available, this early study demonstrated a relationship between paraben toxicity and increasing chain length. Other studies with bacteria have also observed this same trend. Eklund  observed IC50 values in one strain of bacteria ranging from 78 to 837 mg/L, with methylparaben being least toxic and butylparaben being most toxic. These high IC50 values are interesting because parabens are meant to limit microbial growth in many cosmetics, pharmaceuticals, and foods, but the IC50 values are much higher than the fathead minnow and D. magna LC50 and LOEC values observed in the present study. Whereas parabens are more effective against fungi than bacteria , it remains interesting to observe higher effective concentrations in the nontarget species, fathead minnow, and D. magna.
Much attention has been given to antibiotics, especially veterinary antibiotics, in aquatic environments. Results from these studies show a greater chance of acute toxicity from antimicrobial agents such as triclosan and parabens, rather than most human and veterinary antibiotics. Based on the findings of the present study, paraben toxicity to the aquatic species tested is comparable to that of other antimicrobial agents. Triclosan, another antimicrobial agent used in many personal care products, elicited 48 h LC50 values of 0.39 and 0.27 mg/L in D. magna and fathead minnow, respectively. Triclosan, however, exhibited greater toxicity at lower trophic levels including algae and aquatic plant models . Additionally, studies with D. magna exposed to various human and veterinary antibiotics showed LC50 values similar to parabens (oxolinic acid LC50 = 4.6 mg/L), although a number of other LC50 values were orders of magnitude larger (up to 640 mg/L) than paraben LC50s [41–43].
The CTD method utilized in the present study incorporates techniques employed in probabilistic ecological risk assessments, but does not include exposure values, as would be found in a traditional probabilistic risk assessment. Although methods are available to estimate exposure distributions, such as a Monte Carlo analysis using production volumes, no models allow for a probabilistic estimation of environmental concentrations . When there is not sufficient exposure information, the CTD method can be useful in hazard assessment for determining probabilities of finding endpoints or compounds below exposure thresholds based on effect distributions. A laboratory study exposing an aquatic macro-phyte, Lemna gibba, to 12 antibiotics demonstrated the usefulness of this technique in characterizing the hazards associated with a group of compounds that had limited environmental exposure data. The study estimated that 5% of antibiotics would elicit a 10% reduction in wet mass of L. gibba (EC10) at 6 μL. Additionally, it was estimated that 0.3% of antibiotics would adversely affect wet mass of L. gibba at 0.1 μL (the expected environmental concentration) . More recently, Dobbins et al.  demonstrated the utility of CTDs for predicting estrogenicity using several common in vitro and fish models of estrogen agonist activity.
The HQ approach is useful in determining the hazard posed by an individual compound when both toxicity data and exposure data (or predicted exposure values) are available. The CTD approach allows for the hazard associated with an entire group of compounds to be assessed when a representative distribution of effect data for compounds in that class are available. The paraben CTD estimates were derived from distributions of paraben toxicities and were considered to be representative of all parabens, though hazard/risk cannot be absolved from all parabens due to uncertainty. Though, given the low HQ values and slight probability of any parabens eliciting effects at or below 0.1 μL, the hazard to D. magna and larval fathead minnow appear to be minimal. All CTDs predicted a less than one tenth of a percent chance of encountering a paraben in the environment that would elicit an acute or chronic response at a predicted environmental concentration of 0.1 μL. The 5th centile values for mortality and reproduction estimated by the CTD approach are greater than observed concentrations of parabens in surface water and effluent (paraben concentrations ranging from <0.01 to 0.4 μL), while 5th centile values for both D. magna and fathead growth both fall within the range of observed paraben concentrations. However, paraben concentrations have only been determined at sites in developed areas. In less developed countries, wastewater treatment is not as advanced or efficient. In addition, some locations (e.g., India) may have elevated concentrations of contaminants in effluent where some antibiotics were recorded up to a concentration of 31 mg/L . Paraben concentrations will certainly vary at different sites, and more exposure information is needed to further determine whether parabens present hazards to aquatic life.
Whereas there appears to be limited hazard to aquatic organisms exposed to parabens in the environment based on the data collected in the present study, there is potential for parabens and their metabolites to elicit estrogenic responses in aquatic organisms at concentrations lower than those that may cause acute or chronic toxicity. This is especially true for the more lipophilic compounds such as butyl- and benzylparaben. Studies in vitro with MCF-7 cell lines and yeast estrogen screening assays demonstrated estrogenic responses from parabens in the low milligram per liter concentrations [15,17,18]. The full estrogenic potential to aquatic organisms after exposure to parabens has been understudied with few studies examining fish models. Propylparaben elicited vitello-genin induction in Japanese medaka at 9.9 mg/L, but lower concentrations were not tested . Additionally, butylparaben has been shown to induce plasma vitellogenin production in male brown trout at 134 μL and higher . Although studies have not observed estrogenic effects at levels that are at concentrations that have been observed in effluent or surface waters, exposure values are limited in number and in types of sites sampled. Thus, further studies are needed that fully examine the estrogenic potential of all parabens in aquatic ecosytems, particularly to those organisms inhabiting municipal effluent-dominated systems, which represent worst case scenarios for personal care product exposure 
This research was supported by a U.S. Environmental Protection Agency Greater Research Opportunities fellowship to L.L. Dobbins (grant F07E60219), a U.S. EPA grant (grant EM96638001) to B.W. Brooks, and the Baylor University Department of Environmental Science.
- 11984. Final report on the safety assessment of methylparaben, ethylparaben, propylparaben, and butylparaben. J Am Coll Toxicol 3: 147–209..
- 2European Cosmetic Directive. 2007. Cosmetic Directive 76/768/ EEC. Council Directive. Annex VI, Part 1. 1976L0768-EN-19.09.2007–018.002. European Commission. Brussels, Belgium.
- 3Federal Register. 2004. Code of Federal Regulations. Fed Reg 21:184.1490.
- 4Federal Register. 2004. Code of Federal Regulations. Fed Reg 21:184.1670.
- 21U.S. Environmental Protection Agency. 2002. Methods for measuring the acute toxicity of effluents and receiving waters to freshwater and marine organisms, 5th ed. EPA-821-R-02–012. Office of Water, Washington, DC.
- 22U.S. Environmental Protection Agency. 2002. Short-term methods for estimating the chronic toxicity of effluents and receiving waters to freshwater organisms, 4th ed. EPA-821-R-02–013. Office of Water, Washington, DC.
- 23U.S. Environmental Protection Agency. 1994. 10-day chronic toxicity test using Daphnia magna or Daphnia pulex. SOP 2028. Environmental Response Team. Compendium of ERT Standard Operating Protocols. Office of Solid Waste and Emergency Response, Edison, NJ.
- 361997. Predicting modes of toxic action from chemical structure: Acute toxicity in the fathead minnow(Pimephales promelas). Environ Toxicol Chem 16: 984–967., , , , .
- 471995. Exploring QSAR: Hydrophobic, Electronic, and Steric Constants. American Chemical Society, Washington, DC., , .
- 481992. AQUASOL Database of Aqueous Solubility, 5th ed. University of Arizona, College of Pharmacy, Tucson, AZ, USA., .