Screening for PBT Chemicals among the “Existing” and “New

Apr 12, 2012 - In general, the applicability domain of EPI Suite and ECOSAR is ..... the Registration, Evaluation, Authorisation and Restriction of Ch...
1 downloads 0 Views 952KB Size
Download PDF
Policy Analysis pubs.acs.org/est

Screening for PBT Chemicals among the “Existing” and “New” Chemicals of the EU Sebastian Strempel, Martin Scheringer,* Carla A. Ng, and Konrad Hungerbühler Institute for Chemical and Bioengineering, ETH Zürich, Wolfgang-Pauli-Strasse 10, 8093 Zürich, Switzerland S Supporting Information *

ABSTRACT: Under the European chemicals legislation, REACH, industrial chemicals that are imported or manufactured at more than 10 t/yr need to be evaluated with respect to their persistence (P), bioaccumulation potential (B), and toxicity (T). This assessment has to be conducted for several 10 000 of chemicals but, at the same time, empirical data on degradability, bioaccumulation potential and toxicity of industrial chemicals are still scarce. Therefore, the identification of PBT chemicals among all chemicals on the market remains a challenge. We present a PBT screening of approximately 95 000 chemicals based on a comparison of estimated P, B, and T properties of each chemical with the P, B, and T thresholds defined under REACH. We also apply this screening procedure to a set of 2576 high production volume chemicals and a set of 2781 chemicals from the EU’s former list of “new chemicals” (ELINCS). In the set of 95 000 chemicals, the fraction of potential PBT chemicals is around 3%, but in the ELINCS chemicals it reaches 5%. We identify the most common structural elements among the potential PBT chemicals. Analysis of the P, B, and T data for all chemicals considered here shows that the uncertainty in persistence data contributes most to the uncertainty in the number of potential PBT chemicals.

■

INTRODUCTION One of the important innovations of the European chemicals regulation on the registration, evaluation, authorization, and restriction of chemicals (REACH) is that it requires an assessment of chemicals with respect to their persistence, bioaccumulation potential, and toxicity (PBT assessment).1 Criteria for the evaluation of a chemical regarding its P, B, and T properties and also regarding vP (very persistent) and vB (very bioaccumulative) properties are defined in Annex XIII of the REACH regulation. Another important aspect of REACH is that the distinction between “existing” and “new” chemicals has been removed. New chemicals are substances introduced to the market between 1982 and 2007; existing chemicals are substances introduced before 1982. Given the current regulatory focus on PBT chemicals, an interesting question is whether in the new chemicals a trend toward fewer PBT chemicals is visible. The separate listing of new chemicals in the European List of Notified Chemical Substances (ELINCS) provides an opportunity to investigate this question. Before REACH entered into force in 2007, PBT assessment schemes had been established by various institutions, including the OSPAR Convention for the protection of the marine environment of the North-East Atlantic,2 the U.S. Environmental Protection Agency (USEPA),3 and the Stockholm Convention on Persistent Organic Pollutants.4 Also in the earlier Technical Guidance Document on Risk Assessment of the EU, a PBT assessment was foreseen for chemicals with a potential to cause long-term pollution of remote marine ecosystems.5 However, in contrast to these approaches, which © 2012 American Chemical Society

focus on relatively small sets of particularly hazardous compounds (on the order of tens to hundreds of chemicals), REACH is intended to identify all PBT chemicals in the large set of several 10 000 chemicals on the market in the EU. A particular challenge for PBT assessment under REACH is the lack of empirical data on the degradability, bioaccumulation and toxicity of industrial chemicals. Accordingly, key questions related to the PBT assessment under REACH are how many PBT chemicals are to be expected and what types of chemical structures will be present in this set? What are effective approaches to identifying these chemicals? One approach to answering these questions is to define a large set of chemical structures, fill chemical property data gaps with estimated data, screen this database for PBT substances and assess the influence of data uncertainties. Below we briefly describe earlier approaches that deal with screening for PBT chemicals. The USEPA Waste Minimization Prioritization Tool (WMPT)6 combines a scoring algorithm with expert judgment on data quality to indicate levels of concern for the persistence, bioaccumulation potential and toxicity of some 4000 chemicals included in the tool’s internal database. The method was used to develop the PBT list under the USEPA’s Draft Resource Conservation and Recovery Act (RCRA) in 1998; this list has meanwhile been replaced by a list of Priority Chemicals.7 Received: Revised: Accepted: Published: 5680

January 20, 2012 April 6, 2012 April 12, 2012 April 12, 2012 dx.doi.org/10.1021/es3002713 | Environ. Sci. Technol. 2012, 46, 5680−5687

Environmental Science & Technology

Policy Analysis

Pavan and Worth8 used a list of 87 potential PBT chemicals compiled by an EU working group and estimated the PBT characteristics of these chemicals using the estimation tools from the Estimation Program Interface Suite (EPI Suite)9 and ECOSAR;10 in this set, they found 10 chemicals that exceeded the REACH thresholds for P, B, and T. Walker and Carlsen11 used estimation methods from Epi Suite to identify 56 chemicals with potential for high persistence and bioconcentration in a set of 8511 chemicals. Four other studies investigated larger sets of chemicals. Brown and Wania12 developed a method for screening chemicals for long-range transport to the Arctic and bioaccumulation in humans based on partitioning properties and a structural profile derived from 86 known Arctic contaminants. They used the SMILECAS database of 105 584 chemicals13 and identified 822 possible Arctic contaminants of which 120 were high-production volume chemicals. Brown and Wania avoided subjective judgment by applying the same criteria and procedure to all chemicals, but did not address PBT substances directly. Two screening exercises by Muir and Howard14 and Howard and Muir15 focused on chemicals with P and B properties mainly in the Canadian Domestic Substances List (DSL);16 the second study15 also included the USEPA Toxic Substances Control Act (TSCA) Inventory Update Rule (IUR) database.17 Howard and Muir15 used a combination of expert knowledge and P and B criteria from various legislations to select chemicals for further analysis and finally identified 610 chemicals that matched their priority criteria; toxicity was also evaluated but not used as a criterion in the selection of the 610 priority substances. Rorije et al. 18 screened 65 000 industrial chemicals, pharmaceuticals, pesticides, and biocides using a PB score based on overall persistence and bioaccumulation factor; they used their score to rank the chemicals and found 1986 of them exceeded thresholds for POP-like and vPvB substances, whereas 4541 would have been classified as PB based on the PBT criteria they selected from REACH guidelines. Here we substantially expand the scope of earlier screening approaches by applying a set of formally defined P, B, and T criteria, similar to those used by Pavan and Worth,8 to a large set of approximately 95 000 chemicals derived from an extensive processing of chemical structures and property data. To apply the PBT scoring system to these 95 000 compounds, we combine measured PBT property data from multiple sources with estimates for missing property data using methods from EPI Suite and ECOSAR. On this basis, we identify the percentage of potential PBT chemicals among the set of 95 000 chemicals and a subset of high-production volume chemicals (HPVCs); we also discuss the structural fragments that occur frequently in the potential PBT chemicals and the uncertainties of the number of potential PBT chemicals that is associated with using estimated property data. Finally, we evaluate for the first time the number of potential PBT chemicals in a set of “new chemicals” from ELINCS in comparison to the set of chemicals that were on the market as “existing chemicals”.

(EINECS)19 and the SMILECAS database13 we extracted 132 119 CAS registry numbers and SMILES (Simplified Molecular Input Line Entry System) codes. In addition, we collected CAS numbers and SMILES for 5138 chemicals that were introduced on the European market as “new chemicals”, that is, after 1981, and are contained in ELINCS.20,21 To identify the chemical structures of the ELINCS chemicals we had to use a database provided by the German Federal Environment Agency under a confidentiality agreement,20 because for most ELINCS chemicals the CAS numbers and structures are treated as confidential. For these chemicals, we only present aggregated results that do not disclose any individual structure. Duplicates and structures with incorrect or ambiguous CAS numbers or SMILES codes were removed as described in the SI, leaving 127 281 chemical structures in our merged database. Wherever possible, PBT screening should be based on empirical chemical property data. For the property data required for a PBT assessment under REACH (biodegradation half-life, bioconcentration factor and aquatic toxicity), we collected as many individual data points as was possible by searching more than 10 public-domain databases. 22−30 However, the number of experimental property data found is remarkably small; it includes microbial degradation half-lives (P) for 216 chemicals; bioconcentration factors (BCF) (B) for 995 chemicals; acute/chronic aquatic toxicity concentrations (T) for 2198/241 chemicals; and log Kow values for 13 349 chemicals. Only a very small fraction of 91 chemicals had a full set of measured data for degradation half-life, BCF and toxicity. In sections 1.6 and 1.7 of the SI we describe in detail how the different databases were searched and how the data retrieved were processed. However, for the large majority of the chemicals, no freely accessible measured property data were found. Therefore, missing property data were estimated with tools from EPI Suite (Version 4.10):9 BIOWIN3 for half-lives of ultimate biodegradation under aerobic conditions, BCFBAF for BCF, and ECOSAR for 96 h EC50 or LC50 for fish and 48 h EC50 or LC50 for daphnia. In general, the applicability domain of EPI Suite and ECOSAR is given as organic chemicals.9 Therefore, we removed from the data set 20 854 inorganic substances, metallorganic substances and salts by checking the SMILES codes for the presence of metals and charged species. In addition, 1487 carboxylic and sulfonic acids were removed because their BCF derived from BCFBAF was below 2000, but the BCF derived from the Kow estimated for the neutral species was above 2000 (no consistent B classification possible). Next, we removed all chemicals from the database for which the BIOWIN3 output indicated that only the molecular weight was used to estimate biodegradation half-lives (if half-life estimates are based on molecular weight only, the performance of BIOWIN3 is reduced31). In addition, we removed all chemicals with a molecular weight above 1000 g/mol, because such chemicals are outside the ECOSAR domain. Finally, for the 7783 chemicals for which no ECOSAR estimates are available, the LC50 for fathead minnow was estimated with the relationship log LC50 = −log Kow + 2 (LC50 in mmol/L).32 A particular challenge for chemicals assessment are substances of Unknown or Variable composition, Complex reaction products or Biological materials (UVCBs). A systematic treatment of UVCBs is beyond the scope of this work, but our database includes several petroleum-related UVCBs whose CAS represent single chemical structures selected as relevant

■

MATERIALS AND METHODS Chemical Database. The procedure used to compile the set of chemicals included in our PBT screening mostly draws on publicly available data sources and is illustrated in Figure S1 in the Supporting Information (SI). From the European Inventory of Existing Commercial Chemical Substances 5681

dx.doi.org/10.1021/es3002713 | Environ. Sci. Technol. 2012, 46, 5680−5687

Environmental Science & Technology

Policy Analysis

Figure 1. PBT scores of chemicals in the entire set (A), of HPVCs (B), and of ELINCS chemicals (C).

components of the mixture, and also several representative structures for short-chain chlorinated paraffins (SCCPs), another UVCB type. We also removed 473 individual congeners of polychlorinated biphenyls (PCBs), polychlorinated naphthalenes (PCNs), polybrominated diphenyl ethers (PBDEs), polychlorinated dibenzodioxins and -furans (PCDD/ Fs) and represent these substances as homologues in our database. All these steps are described in detail in section 1 of the SI. After these steps, 94 483 chemicals remained in our database. Of these, 2576 are listed as HPVCs by either the OECD or the USEPA, 2781 are contained in ELINCS, and 49 498 have been preregistered under REACH.33 PBT Scoring. To characterize the chemicals in the database according to their PBT properties, we developed a PBT score based on the P, B, and T thresholds defined under REACH. For each chemical, the score aggregates equally weighted subscores for P, B, and T. For each of these hazard dimensions we used a chemical property as required by Annex XIII of REACH: biodegradation half-life, BCF and chronic NOEC or acute LC50 or EC50. These property data were compared to the following threshold values: for persistence, a biodegradation half-life of 120 days in soil; for bioconcentration, a BCF of 2000; and for toxicity a chronic NOEC of 0.01 mg/L or an acute effect concentration of 0.1 mg/L. These thresholds were chosen on the basis of Annex XIII of the REACH legislation1 and the Guidance on PBT assessment of the European Chemicals Agency.34 For each property we calculated the subscore as the ratio between the property value and the threshold. To avoid the situation that one hazard property of a substance has such a large subscore that it would dominate the aggregated score and hide information on the other two properties, we truncated each subscore exceeding the value of 1 at 1.0. To obtain the final PBT score for each substance, we divided the three subscores by three and added them up. We also developed a corresponding scheme for vPvB substances, see SI. PBT scores in the interval [0, 0.333) represent nonpersistent, nonbioconcentrating, and nontoxic chemicals for which the score directly shows that no threshold is exceeded. On the other hand, a PBT score of 1.0 indicates that all three thresholds are exceeded and the chemical may be a PBT substance. Scores in the interval [0.333, 1) may be composed of different combinations of subscores. For instance, a score of 0.4

may be composed of one subscore that exceeded 1 and was truncated at 1.0 and two subscores of 0.1 (1.0 + 0.1 + 0.1)/3 = 0.4) or be composed of three subscores below the thresholds, for example, (0.3 + 0.4 + 0.5)/3 = 0.4. The PBT score calculated in this way does not indicate by how much a chemical may exceed a threshold. Therefore, the score cannot be used to rank the chemicals, but it can be used to assign them to four hazard classes: the class “PBT” contains all chemicals with a score of 1.0, class “nonPBT2” all chemicals with two subscores equal to 0.333 (exceeding two thresholds), class “nonPBT1” all chemicals with one subscore equal to 0.333 (exceeding one threshold), and class “nonPBT0” all chemicals with all three subscores below 0.333 (no threshold exceeded). Sensitivity and Uncertainty Analysis. To assess the sensitivity of the number of potential PBT chemicals to changes in the property data, we varied the toxicity, persistence, and bioconcentration factors of all chemicals by a factor of 2 around the best property estimates, and evaluated how many chemicals were added to or removed from the PBT hazard class. To assess the uncertainty of the chemical property data, we used estimated and measured property data in combination. For the chemicals for which measured data could be retrieved, we plotted the estimated against the measured data (separately for half-lives, bioconcentration factors, acute and chronic toxicities, and octanol/water partition coefficients) and derived uncertainty factors from the scatter in these plots, as described in section 1.7.1 of the SI. We then applied these uncertainty factors in our screening scheme to determine the impact of uncertain property data on the fraction of chemicals in the PBT class.

■

RESULTS Potential PBT Chemicals. We calculated PBT scores for the entire database and classified each of the substances according to the four hazard classes, see Figure 1A. There are 2930 substances in the PBT class (3.1%), 9302 nonPBT2 substances (9.84%), 24 320 nonPBT1 substances (25.7%), and 57 931 chemicals classified as nonPBT0 (61.3%). For the 2576 HPVCs and the 2781 chemicals from ELINCS, Figure 1B and C display the distributions of the PBT scores. The colors indicate the four hazard classes, ranging from lightest (nonPBT) to darkest (PBT). In addition to showing the fractions of chemicals in each class, the histograms make it possible to visually compare the distribution of the scores in the

5682

dx.doi.org/10.1021/es3002713 | Environ. Sci. Technol. 2012, 46, 5680−5687

Environmental Science & Technology

Policy Analysis

diphenylethers, phenylamines, and diphenylamines, in some cases also polyphenols and polyphenylamines, with several highly branched substituents. These compounds are, for example, used as UV absorbers (antidegradants) in synthetic rubber. We found 13 such compounds; three of them were also identified by Howard and Muir. 15 Another group is halogenated compounds, including chlorinated paraffins, hexabromocyclododecane, pentabromo diphenylether, and tetrabromophthalic anhydride (flame retardants), and tetrabromopyrene (a chemical intermediate for the production of electroluminescent displays) and several compounds with perfluorinated alkyl chains. Finally, highly branched alkyl compounds are found, such as heptamethyl nonane. PBT Scores of Chemicals from ELINCS. The PBT scores of the 2781 chemicals that were listed as new chemicals in the EU are shown in Figure 1C. Notable is the fraction of 5.2% of chemicals in the PBT class, which is even higher than in the entire data set and in the HPVC set. This finding indicates that in the set of approximately 5500 chemicals that were introduced on the European market between 1982 and 2007 there is no tendency toward a significantly lower proportion of PBT chemicals. Compared to the potential PBT chemicals in the entire set, there is a shift from brominated and chlorinated substances toward fluorinated substances in the potential PBT chemicals from ELINCS: brominated chemicals decrease from 9.1% in the potential PBTs of the entire set to only 1.4% of the potential PBTs in ELINCS; the fraction of chlorinated potential PBTs similarly decreases from 31% to 13%. By contrast, the fraction of potential PBTs containing fluorine in the ELINCS set is 29%, compared to 14% in the full set. These chemicals are not poly- or perfluorinated alkyl substances such as PFOA, but contain other fluorinated moieties. PBT Scores of Some Restricted Chemicals and Their Replacements. It is interesting to compare the PBT scores of some currently restricted chemicals, for example, PCBs and PBDEs, to those of their existing and potential substitutes, namely SCCPs in the case of PCBs and various brominated aromatic compounds in the case of PBDEs. Of the PCBs, the dichloro homologues do not exceed the P threshold; all other homologues are in the PBT class. As for the PBDEs, the dibromo homologue is below the P threshold and the hepta- to decabromo homologues are below the B threshold. The tri- to hexabromo homologues are in the PBT class. In the EU SCCPs were used as replacements for PCBs,38 but are now on the candidate list for Substances of Very High Concern; their use has been restricted in many EU countries. In other parts of the world, however, SCCPs are still used in amounts of several 100 000 t/yr as replacements of PCBs39 and levels measured in the environment exceed those of PCBs.38 In our database, SCCPs are represented by six lead substances with chlorine contents from 40% to 61%, see list in the SI. Four of these (chlorination degree: 46%, 54%, 60%, and 61%) are in our PBT class; one is in nonPBT2 (51% chlorine, P and T exceeded); one is in nonPBT1 (40% chlorine, T exceeded). PBDEs might be replaced by other brominated substances including, for example, 1,2-bis(pentabromophenyl) ethane (DBDPE, CAS 84852-53-9), 1,2-bis(2,4,6-tribromophenoxy) ethane (BTBPE, CAS 37853-59-1), ethylene bis-tetrabromophthalimide (EBTPI, CAS 32588-76-4), dibromopropyl-tribromophenyl ether (DPTE, CAS 35109-60-5), pentabromotoluene (PBTo, CAS 87-83-2), pentabromoethylbenzene (PBEB, CAS 85-22-3), and hexabromobenzene (HxBBz, CAS 87-82-

three sets of chemicals. Note that for PBT scores between 0.3 and 0.999 the bars of the higher classes (darker colors) are stacked. That is, they begin at the top of the lighter bars for the next lower class, not at the x-axis. The dark red bar for the potential PBT chemicals extends from 1.0 to 1.05 for graphical reasons; all chemicals represented by this bar have a score of 1.0. The distributions for the classes nonPBT2, nonPBT1, and nonPBT0 overlap to some extent. This is because all three characteristics (P, B, and T) contribute to the total score, thus a chemical exceeding only one threshold (nonPBT1) can still have the same total score as a chemical exceeding two thresholds (nonPBT2). We screened the 2930 chemicals in the PBT class for the most common structural elements. Most frequent elements are chlorinated and brominated aromatic systems (benzenes, naphthalenes, biphenyls, diphenylethers, dibenzodioxins, and -furans); chlorinated and brominated cycloaliphatic compounds, including, for example, the chlorinated norbornene structure of the hexachloro cyclopentadiene insecticides; highly branched alkyl substances and aromatic compounds (often phenols and phenylamines) with several highly branched alkyl, ether, or tertiary amine groups as substituents; triphenylmethyl substances; spiro compounds; various per- and polyfluorinated alkyl substances of different chain lengths; compounds with trifluoromethyl substituents; nitroaromatic compounds; tertiary amines with highly branched alkyl groups; and polycyclic aromatic hydrocarbons. Also combinations of two or more of these structural elements are frequently found in the PBT hazard class. The number of chemicals in the PBT class that do not contain F, Cl, or Br is 1509 (52% of the chemicals in the PBT class). A list of the CAS numbers and SMILES codes of the 2930 chemicals in the PBT class (except for 143 from ELINCS) is provided in the SI. As of March 2012, 32 of the 2930 potential PBT chemicals were registered under REACH35 (indicated in Table S11 in the SI). There are 1202 potential vPvB substances (1.3%) in the database. Of these, 1201 are also in the PBT hazard class. This high fraction of potential vPvB chemicals that are also potentially PBT indicates the known correlation of the B and T dimensions.36,37 This correlation is caused by the fact that chemicals with high B scores partition strongly into lipid tissue and exert a corresponding level of baseline toxicity.32 To investigate this relationship for the 2930 chemicals in the PBT class, we estimated their baseline toxicity (LC50,bl) using the relationship log LC50,bl = −log Kow + 2 for LC50,bl in units of mmol/L32 and compared these LC50,bl values to the T threshold of 0.1 mg/L. For 2891 of the 2930 chemicals in the PBT class, the LC50,bl value was below 0.1 mg/L, which means that 98% of the chemicals in the PBT class could have been identified as T chemicals on the basis of their estimated baseline toxicity alone. PBT Scores of HPV Chemicals. Figure 1B shows the distribution of the PBT scores of the 2576 HPV chemicals. This distribution is similar to the distribution of the chemicals in the entire database; the percentages (numbers) of HPVCs in the four hazard classes are: 2.2% (57) in PBT, 8.4% (216) in nonPBT2, 28.9% (744) in nonPBT1, and 60.6% (1559) in nonPBT0. The 57 chemical structures with a PBT score of 1.0 are listed in Table S5 of the SI; they include 24 petroleumbased products such as hydrocracked petroleum residues and aromatic hydrocarbons with more than 20 carbon atoms. Uses of these petroleum-based compounds include, for example, lubricating oils. The remaining 33 chemicals include phenols, 5683

dx.doi.org/10.1021/es3002713 | Environ. Sci. Technol. 2012, 46, 5680−5687

Environmental Science & Technology

Policy Analysis

1),40 see SI. One of these seven compounds, DPTE (CAS 35109-60-5), is in our PBT class; the other six are in class nonPBT2 (P and T thresholds exceeded). Sensitivity and Uncertainty Analysis. To evaluate the sensitivity of our screening scheme to changes in the chemical properties, we varied eachP, B, and Tby a factor of 2. Figure 2 shows the distributions of the three individual

As described in section 1.7.1 of the SI, our analysis of the uncertainty in the different property data yielded the following uncertainty factors: 4 for half-life, 4 for BCF, 100 for chronic toxicity, 45 for acute toxicity, and 3.5 for Kow. Applying these uncertainty factors to our screening of the entire database yields a minimum of 153 chemicals classified as potential PBTs (0.26%) and a maximum of 12 493 potential PBTs (13.2%), as compared to our initial estimate of 2930 (3.1%). Although the uncertainty range of the half-lives is similar to that of BCF and Kow and substantially smaller than for toxicity, it nevertheless still contributes most to this range, as expected from the sensitivity analysis described above. Thus, in order to more accurately estimate the number of potential PBT chemicals, it will be crucial to obtain better half-life data.

■

DISCUSSION Our results stem from a screening exercise on a very large number of chemicals. Overall, the “big picture” that emerges from our results is plausible and leads to several interesting insights. A general caveat in the context of such screening exercises (including also similar studies conducted earlier12,14,15,18) is that, for individual chemicals, errors in the chemical property estimates and the P, B, and T characteristics are possible. For this reason, we call the chemicals for which we obtained a PBT score of 1.0 “potential PBT chemicals”. Regarding the performance of ECOSAR, it is known that ECOSAR mainly derives toxicity estimates from the Kow and may not perform well for chemicals with specific modes of action.41 This limitation is most important for polar and reactive chemicals with low Kow, because their specific toxicity may well exceed their baseline toxicity, which is related to the Kow.37 For these chemicals, the ECOSAR estimates and, correspondingly, our T scores may be systematically too low. However, this limitation is much less problematic for chemicals with high P and B, because these chemicals in any case exert a strong baseline toxicity. Because we intend to identify these chemicals, the general limitation of ECOSAR does not strongly affect our estimate of the number of potential PBT chemicals. In addition, it may be possible that BCFBAF does not identify chemicals that bioaccumulate because of binding to proteins. For these chemicals, our estimated B subscores may be too low. Where available, we took information on applicability domains of the estimation methods into account; according to this information, we removed 32 325 chemicals from the initial list of 126 808 chemicals with valid SMILES codes. It is important to note that the results of our analysis can be refined as soon as more specific (empirical) information on degradation half-lives, BCF and toxicity of a chemical becomes available. To evaluate our screening system, we can consider the PBT scores for three relatively well characterized groups with high fractions of known PBT chemicals: PCBs, PCDD/Fs, PBDEs. The results obtained for these three groups are in agreement with what can be expected: highly chlorinated compounds are all in the PBT class (except octachloro dibenzofuran, which is nonPBT2 because of a relatively low B score). For PBDEs the situation is somewhat different, because here the B scores are below the threshold for the highly brominated compounds (hepta to deca). For deca- and nona-BDE, the estimated BCFs are particularly low, which is confirmed by the stringent method for identifying non-B chemicals developed by Nendza

Figure 2. Distributions of subscores for P (top), B (middle), and T (bottom). In each panel, a value of 0.333 indicates the REACH threshold; dashed vertical lines and colors indicate the range defined by a factor of 2 around the threshold.

subscores on the entire set of 94 483 chemicals. The dashed vertical lines indicate, for each distribution, how many chemicals are removed from (orange region) or added to (green region) the PBT class if the respective property is multiplied or divided by 2. The results in Figure 2 show that the same change in the P, B, or T data has different effects on the number of potential PBT chemicals. The biggest contribution to changes in the classification of chemicals as potential PBTs stems from the half-life data. This is because there are many more chemicals with half-lives close to the P threshold than chemicals that are close to the B and T thresholds. 5684

dx.doi.org/10.1021/es3002713 | Environ. Sci. Technol. 2012, 46, 5680−5687

Environmental Science & Technology

Policy Analysis

and Müller.42 In contrast, the octa-BDE and hepta-BDEs might actually exceed the B threshold, because their estimated BCFs are between 850 and 1930, which is within the uncertainty factor of 4 around the threshold of 2000. In general, the structural fragments found in the 2930 chemicals in the PBT class are in agreement with established knowledge on chemical properties that lead to P and B characteristics: these are various recalcitrant structural elements43 and log Kow values between 4 and 8.44 When our results are compared to the PBT chemicals identified by the USEPA,45 there is substantial overlap (in spite of some differences in the thresholds used by the EPA relative to REACH). Of the 64 PBT chemicals listed by the USEPA,45 39 are in our database (the 25 chemicals of the EPA list that are not in our database are individual PCB and PCDD/F congeners that were removed from our database). Of the 39 chemicals in common, 27 are in our PBT class; 10 are in nonPBT2 because of B scores below the threshold (examples: chrysene, methoxychlor, octachloro dibenzofuran), the remaining two are in nonPBT1 because of P and B scores below the thresholds; for a list of all 39 and their individual subscores, see SI. For both B and T, the EPA thresholds are less stringent than under REACH, that is, include more chemicals: the EPA threshold for BCF is 1000 (2000 in REACH), and the EPA toxicity thresholds range from 10 mg/L (moderate concern) to 0.1 mg/L (high concern). Using the EU list of 125 potential PBT substances, Pavan and Worth8 determined P, B, and T scores for 87 compounds (the remaining 38 could not be assessed because they are mixtures or polymers). They found 10 chemicals that exceed the P, B, and T thresholds of REACH; eight of these are also in our PBT class. Rorije et al.18 in their screening of 65 000 industrial chemicals, pharmaceuticals and pesticides and biocides identified 7% of potential PBT chemicals and 3% of potential vPvB chemicals, which is higher than our fractions of 3% (PBT) and 0.84% (vPvB). For all chemicals, they used estimated property data; they did not include any chemicals from ELINCS. Brown and Wania12 identified chemicals with Arctic Contamination Potential (ACP), which is related, but not identical to the PBT characteristics of a chemical. Of the 822 chemicals with high ACP identified by Brown and Wania, 519 are in our database and 130 of these are in our PBT class, see SI. Several types of chemicals in our PBT class are the same as identified by Howard and Muir,15 including various brominated flame retardants, tetrabromopyrene, several poly- and perfluorinated chemicals, and several UV absorbers, see SI. For the priority chemicals identified by Brown and Wania12 and Howard and Muir,15 the fractions of chemicals in the four hazard classes are shown in Figure 3, along with the same fractions for three additional sets of chemicals: the SIN list (Substitute It Now) of the International Chemical Secretariat,46 the ETUC list,47 and the OSPAR List of Substances of Possible Concern.48 The fraction of chemicals in the PBT class varies considerably among these five lists of priority chemicals, which shows that the criteria of concern underlying the lists differ and do not necessarily focus on PBT properties. A key finding of our analysis is that there is a fraction of around 3% of potential PBT chemicals in the entire set of chemicals. Many of these chemicals have probably never been

Figure 3. Fractions of chemicals in the four hazard classes. BW: chemicals with high ACP of Brown and Wania; HM: priority chemicals of Howard and Muir; OSP: OSPAR Convention; SIN: “Substitute it Now”, International Chemical Secretariat; ETUC: European Trade Union Confederation. Numbers at the bottom: number of chemicals included in each bar (overlap between the different sets and our database).

investigated for potential PBT properties. If we apply the uncertainty factors we have determined for the individual chemical properties (half-lives, BCF and toxicity data) to propagate uncertainty through our assessment scheme, the fraction of potential PBT chemicals ranges from 0.16% to 13.2%. The largest part of this uncertainty range is caused by the uncertainty in degradation half-lives, which implies that more and more accurate degradation half-life data are essential for a better identification of PBT chemicals. However, even in the light of the current uncertainties, our results imply that it is likely that there are several hundred PBT chemicals on the market. During the registration process under REACH, it will be important to better characterize the use patterns of these chemicals, to estimate their emissions into the environment, and to obtain better chemical property data. Our analysis further shows that the vPvB class under REACH makes sense from a regulatory point of view, because most vPvB chemicals also have high T scores just because of their baseline toxicity, which is directly related to their high BCF. In other words, by focusing on P and B as two independent properties, the vPvB class makes it possible to reduce the efforts for toxicity testing. On the other hand, chemicals that have a PBT score below 0.333 (53%, 50 740 chemicals) are probably not particularly hazardous to the environment in terms of PBT properties and could be candidates for waiving of experimental testing as foreseen in REACH. A second key finding is the surprisingly high number of potential PBT chemicals among the chemicals from ELINCS (5.2%). ELINCS contains about 5500 chemicals that were introduced to the EU market between 1982 and 2007, and here we present the first estimate of the number of potential PBT chemicals in this set. From our analysis, it appears that the fraction of potential PBT chemicals in the newer chemicals is even higher than in our entire set of chemicals, 97% of which are existing chemicals. In other words, no trend toward an elimination of chemicals with inherent environmentally hazardous properties is visible in the PBT characteristics of the new chemicals introduced to the market between 1982 and 2007. This leads to the question of how the concept of green design of chemical products or “benign by design”49 can be made more effective in the development of new chemical products. 5685

dx.doi.org/10.1021/es3002713 | Environ. Sci. Technol. 2012, 46, 5680−5687

Environmental Science & Technology

■

However, it is important to note that, before the entry-intoforce of REACH, chemicals legislation in the EU did not focus on PBT properties so that there was no clear incentive to avoid PBT properties in chemical products. Our results illustrate that the stronger emphasis on PBT assessment under REACH is justified. Because of the high number of chemicals to be assessed, PBT screening methods will play an important role in the implementation of REACH. A related problem arises from the substitution of PCBs and PBDEs, which to a large extent are PBT chemicals, by SCCPs and by brominated flame retardants other than PBDEs: the substitutes are very similar to the initial compounds in both their desirable and their undesirable properties and they are likely to be PBT chemicals as well. A similar observation can be made for poly- and perfluorinated chemicals, where compounds with eight carbon atoms (perfluorooctanoic acid, PFOA, and perfluorooctane sulfonic acid, PFOS and related compounds) are replaced by analogues with shorter carbon chains,50 which are as persistent and in some cases as bioaccumulative as the C8 compounds.15 These observations indicate that (i) established chemical products, especially when they are technically and economically important, are often substituted by chemicals with similar unfavorable environmental properties and (ii) the development of new chemical products is not yet directed toward significantly less persistent (and bioaccumulative and toxic) chemicals. Degradability or low persistence has been emphasized since the 1970s as the key to more environmentally benign chemical products.51−53 Our results indicate that the goal of less persistent chemical products has not been reached in the last decades of chemical product development. Concerning the contribution of Green Chemistry to the development of a new generation of environmentally benign products, we conclude from our analysis of both “existing” and “new” chemicals that it is of utmost importance that PBT properties (and degradability in particular) are considered as performance indicators from the very beginning of the product development process.

■

REFERENCES

(1) Regulation (EC) No 1907/2006 of the European Parliament and of the Council of 18 December 2006 concerning the Registration, Evaluation, Authorisation and Restriction of Chemicals (REACH), establishing a European Chemicals Agency, amending Directive 1999/ 45/EC and repealing Council Regulation (EEC) No 793/93 and Commission Regulation (EC) No 1488/94 as well as Council Directive 76/769/EEC and Commission Directives 91/155/EEC, 93/67/EEC, 93/105/EC and 2000/21/EC . Off. J. Eur Union 2006, 49, 1−849. (2) Convention for the Protection of the Marine Environment of the North-East Atlantic, 1992. http://www.ospar.org/ (accessed March 30, 2012). (3) Persistent, Bioaccumulative, and Toxic Profiles Estimated for Organic Chemicals On-Line, 2006. http://www.pbtprofiler.net/ (accessed March 30, 2012). (4) Stockholm Convention on Persistent Organic Pollutants (POP), 2001. http://chm.pops.int/ (accessed March 30, 2012). (5) Technical Guidance Document on Risk Assessment in support of Commission Directive 93/67/EEC on Risk Assessment for new notified substances, Commission Regulation (EC) No 1488/94 on Risk Assessment for existing substances, Directive 98/8/EC of the European Parliament and of the Council concerning the placing of biocidal products on the market, Part II, Chapter 4.4; European Communities: Brussels, 2003. (6) Ralston, M. D.; Fort, D. L.; Jon, J. H.; Kwiat, J. K. In The U.S. Environmental Protection Agency Waste Minimization Prioritization Tool: Computerized System for Prioritizing Chemicals Based on PBT Characteristics, Chapter 13; ACS Symposium Series; American Chemical Society: Washington, DC, 2000; Vol. 773. (7) Priority Chemicals, United States Environmental Protection Agency. http://www.epa.gov/osw/hazard/wastemin/priority.htm (accessed March 30, 2012). (8) Pavan, M.; Worth, A. A Set of Case Studies to Illustrate the Applicability of DART (Decision Analysis by Ranking Techniques) in the Ranking of Chemicals; European Commission Joint Research Centre: Ispra, Italy, 2008; ecb.jrc.ec.europa.eu/documents/QSAR/EUR_ 23481_EN.pdf (accessed March 30, 2012). (9) Estimation Programs Interface Suite for Microsoft Windows, v 4.10; United States Environmental Protection Agency: Washington, DC, USA, 2011. http://www.epa.gov/opptintr/exposure/pubs/ episuite.htm (accessed March 30, 2012). (10) Ecological Structure Activity Relationships (ECOSAR), United States Environmental Protection Agency, 2009. http://www.epa.gov/ opptintr/newchems//tools/21ecosar.htm (accessed March 30, 2012). (11) Walker, J. D.; Carlsen, L. QSARs for identifying and prioritizing substances with persistence and bioconcentration potential. SAR QSAR Environ. Res. 2002, 13, 713−725, DOI: 10.1080/ 1062936021000043454. (12) Brown, T.; Wania, F. Screening chemicals for the potential to be persistent organic pollutants: a case study of Arctic contaminants. Environ. Sci. Technol. 2008, 42, 5202−5209, DOI: 10.1021/es8004514. (13) SMILECAS Database, Syracuse Research Corporation. http:// www.srcinc.com/what-we-do/product.aspx?id=135 (accessed March 30, 2012). (14) Muir, D. C. G.; Howard, P. Are there other persistent organic pollutants? A challenge for environmental chemists. Environ. Sci. Technol. 2006, 40, 7157−7166, DOI: 10.1021/ es061677a. (15) Howard, P.; Muir, D. C. G. Identifying new persistent and bioaccumulative organics among chemicals in commerce. Environ. Sci. Technol. 2010, 44, 2277−2285, DOI: 10.1021/es903383a. (16) Domestic Substances List (DSL), Environment Canada. http:// www.ec.gc.ca/lcpe-cepa/default.asp?lang=En&n=5F213FA8-1 (accessed March 30, 2012). (17) United States Environmental Protection Agency; Inventory Updated Reporting under the Toxic Substances Control Act. http:// www.epa.gov/iur (accessed March 30, 2012). (18) Rorije, E. Verbruggen, E. M. J.; Hollander, A.; Traas, T. P.; Janssen, M. P. M. Identifying potential POP and PBT substances: Development of a new persistence/bioacumulation-score. RIVM

ASSOCIATED CONTENT

S Supporting Information *

Detailed description of data sources and data processing; distributions of PBT scores for test chemicals; lists of potential PBT chemicals from several subsets; entire list of potential PBT chemicals. This material is available free of charge via the Internet at http://pubs.acs.org.

■

Policy Analysis

AUTHOR INFORMATION

Corresponding Author

*Phone: +41 44 632 3062; fax: +41 44 632 1189; e-mail: [email protected]. Notes

The authors declare no competing financial interest.

■

ACKNOWLEDGMENTS Funding by the Integrated EU Project OSIRIS (contract GOCE-ET-2007-037017) is gratefully acknowledged. We thank C. Schulte, M. Pritzsche, and S. Helmich (all German Federal Environment Agency) for support with data from ELINCS. 5686

dx.doi.org/10.1021/es3002713 | Environ. Sci. Technol. 2012, 46, 5680−5687

Environmental Science & Technology

Policy Analysis

report 601356001, 2011. http://rivm.nl/bibliotheek/rapporten/ 601356001.html (accessed March 30, 2012). (19) EC Inventory − Institute for Health and Consumer Protection − (JRC − IHCP), European Commission: EINECS. http://ihcp.jrc. ec.europa.eu/our_labs/computational_toxicology/informationsources/ec_inventory (accessed March 30, 2012). (20) Contract for Research between ETH Zurich and the German Federal Environment Agency (Umweltbundesamt, UBA), FKZ 360 01 067; Klassifizierung der PBT-Eigenschaften der ehemaligen Neustoffe durch QSAR-Berechnungen fur die Identifizierung weiterer besonders besorgniserregender Stoffe unter REACH. (21) Regulation (EC) No 1272/2008 of the European Parliament and of the Council of 16 December 2008 on classification, labelling and packaging of substances and mixtures, amending and repealing Directives 67/548/EEC and 1999/45/EC, and amending Regulation (EC) No 1907/2006. (22) EPI Suite Data. http://esc.syrres.com/interkow/EpiSuiteData. htm (accessed March 30, 2012). (23) SRC BIODEG database. http://srcinc.com/what-we-do/ databaseforms.aspx?id=382 (accessed March 30, 2012). (24) Ecological Categorization Results from the Canadian Domestic Substance List, Environment Canada. http://webnet.oecd.org/ cm3rweb/Default.aspx (accessed March 30, 2012). (25) Data bank of Environmental Properties of Chemicals − EnviChem; Finish Environment Institute (SYKE). http://www. ymparisto.fi/default.asp?contentid=141944&lan=en (accessed March 30, 2012). (26) Quantitative Structure Activity Relationships (QSARs), Risk Management Sustainability Research, USEPA. http://www.epa.gov/ nrmrl/std/cppb/qsar/ (accessed March 30, 2012). (27) Dimitrov, S.; Dimitrova, N.; Parkerton, T.; Combers, M.; Bonnell, M.; Mekenyan, O. Base-line model for identifying the bioaccummulation potential of chemicals. SAR QSAR Environ. Res. 2005, 16, 531−554, DOI: 10.1080/10659360500474623. (28) AQUIRE (AQUatic toxicity Information Retrieval); USEPA. http://cfpub.epa.gov/ecotox/ (accessed March 30, 2012). (29) Chemical Risk Information Platform (CHRIP); National Institute of Technology and Evaluation (Japan). http://www.safe. nite.go.jp/english/db.html (accessed March 30, 2012). (30) BCF gold standard database; CEFIC LRI. http://ambit. sourceforge.net/euras/ (accessed March 30, 2012). (31) Boethling, R. S.; Costanza, J. Domain of EPI suite biotransformation models. SAR QSAR Environ. Res. 2010, 21, 415− 443, DOI: 10.1080/1062936X.2010.501816. (32) Maeder, V.; Escher, B. I.; Scheringer, M.; Hungerbühler, K. Toxic Ratio as an indicator of the intrinsic toxicity in the assessment of persistent, bioaccumulative, and toxic chemicals. Environ. Sci. Technol. 2004, 38, 3659−3666, DOI: 10.1021/ es0351591. (33) Pre-Registered Substances; European Chemicals Agency: Helsinki, Finland; http://echa.europa.eu/web/guest/information-on-chemicals/ pre-registered-substances (accessed March 30, 2012). (34) Guidance on Information Requirements and Chemical Safety Assessment, Chapter R.11: PBT Assessment; European Chemicals Agency: Helsinki, Finland, 2008; http://guidance.echa.europa.eu/ docs/guidance_document/information_requirements_r11_en. pdf?vers=20_08_08 (accessed March 30, 2012). (35) Registered Substances.; European Chemicals Agency: Helsinki, Finland; http://echa.europa.eu/web/guest/information-on-chemicals/ registered-substances (accessed March 30, 2012). (36) Verhaar, H. J. M.; Van Leeuwen, C. J.; Hermens, J. L. M. Classifying environmental pollutants. 1: Structure-activity relationships for prediction of aquatic toxicity. Chemosphere 1992, 25, 471−491, DOI: 10.1016/0045-6535(92)90280-5. (37) Escher, B. I.; Hermens, J. L. M. Modes of action in ecotoxicology: Their role in body burdens, species sensitivity, QSARs, and mixture effects. Environ. Sci. Technol. 2002, 36, 4201− 4217, DOI: 10.1021/es015848h. (38) Iozza, S.; Müller, C. E.; Schmid, P.; Bogdal, C.; Oehme, M. Historical profiles of chlorinated paraffins and polychlorinated

biphenyls in a dated sediment core from Lake Thun (Switzerland). Environ. Sci. Technol. 2008, 42, 1045−1050, DOI: 10.1021/es702383t. (39) Fiedler, H. Short-chain chlorinated paraffins: production, use and international regulations. In Chlorinated Paraffins, Handbook of Environmental Chemistry 10; J. de Boer, Ed.; Springer: Berlin Heidelberg, 2010; pp 1−40. (40) Covaci, A.; Harrad, S.; Abdallah, M. A. E.; Ali, N.; Law, R. J.; Herzke, D.; de Wit, C. A. Novel brominated flame retardants: A review of their analysis, environmental fate and behaviour. Environ. Int. 2011, 37, 532−556, DOI: 10.1016/j.envint.2010.11.007. (41) Moore, D. R. J.; Breton, R. L.; MacDonald, D. B. A comparison of model performance for six quantitative structure-activity relationship packages that predict acute toxicity to fish. Environ. Toxicol. Chem. 2003, 22, 1799−1809, DOI: 10.1897/00-361. (42) Nendza, M.; Mü l ler, M. Screening for low aquatic bioaccumulation (1): Lipinski’s ‘Rule of 5′ and molecular size. SAR QSAR Environ. Res. 2010, 21, 495−512, DOI: 10.1080/ 1062936X.2010.502295. (43) Boethling, R. S.; Sommer, E.; DiFiore, D. Designing small molecules for biodegradability. Chem. Rev. 2007, 107, 2207−2227, DOI: 10.1021/cr050952t. (44) Arnot, J. A.; Gobas, F. A. P. C. A review of bioconcentration factor (BCF) and bioaccumulation factor (BAF) assessments for organic chemicals in aquatic organisms. Environ. Rev. 2006, 14, 257− 297, DOI: 10.1139/A06-005. (45) Persistent, bioaccumulative, and toxic organic chemicals in EPA’s final PBT Rule for TRI. http://www.pbtprofiler.net/ methodology.asp (accessed March 30, 2012). (46) SIN (Substitute It Now!) list; http:// http://www.sinlist.org/ (accessed March 30, 2012). (47) ETUC earmarks priority chemicals for REACH authorization; http://www.etuc.org/a/6022 (accessed March 30, 2012). (48) Substances of possible concern − Hazardous Substances − Work Areas − OSPAR Commission; http://www.ospar.org/v_ substances/browse.asp?menu= 00950304450072_000000_000000 (accessed March 30, 2012). (49) Kümmerer, K. Sustainable from the very beginning: rational design of molecules by life cycle engineering as an important approach for green pharmacy and green chemistry. Green Chem. 2007, 9, 899− 907, DOI: 10.1039/b618298b. (50) Ritter, S. K. Fluorochemicals go short. Chem. Eng. News 2010, 88, 12−17. (51) Stephenson, M. S. An approach to the identification of organic compounds hazardous to the environment and human health. Ecotoxicol. Environ. Saf. 1977, 1, 39−48. (52) Klöpffer, W. Environmental hazardAssessment of chemicals and products, Part II: Persistence and degradability of organic chemicals. Environ. Sci. Pollut. Res. 1994, 1, 108−116. (53) Scheringer, M. Persistence and spatial range as endpoints of an exposure-based assessment of organic chemicals. Environ. Sci. Technol. 1996, 30, 1652−1659, DOI: 10.1021/es9506418.

5687

dx.doi.org/10.1021/es3002713 | Environ. Sci. Technol. 2012, 46, 5680−5687