A mixture of: triphenylthiophosphate and tertiary butylated phenyl derivatives

Bioaccumulation: aquatic / sediment

Description of key information

Based on the available data comprising experimental data, QSAR and other factors relevant for bioaccumulation, it can be concluded that Structure A has to be considered bioaccumulative in aquatic organisms in a worst case approach (= bioaccumulation test with concentrations above water solubility). This substance is currently being evaluated in CORAP. A new bioaccumulaion test will be performed to verify to experimental results. The other constituents are considered to have a BCF < 2000.

Additional information

In Article 13 of Regulation (EC) No 1907/2006, it is laid down that information on intrinsic properties of substances may be generated by means other than tests, provided that the conditions set out in Annex XI (of the same Regulation) are met. Furthermore according to Article 25 of the same Regulation testing on vertebrate animals shall be undertaken only as a last resort. According to Annex XI of Regulation (EC) No 1907/2006 (Q)SAR results can be used if (1) the scientific validity of the (Q)SAR model has been established, (2) the substance falls within the applicability domain of the (Q)SAR model, (3) the results are adequate for the purpose of classification and labeling and/or risk assessment and (4) adequate and reliable documentation of the applied method is provided. For the assessment (Q)SAR results were used for aquatic bioaccumulation. The criteria listed in Annex XI of Regulation (EC) No 1907/2006 are considered to be adequately fulfilled and therefore the endpoint(s) sufficiently covered and suitable for risk assessment. Additionally, experimental data with a similar substance (CAS 192268-65-8 containing O,O,O- triphenyl thiophosphate as major constituent) are available. The BCF of the individual constituents was determined. Therefore, and for reasons of animal welfare, further experimental studies on bioaccumulation are not provided. The bioaccumulative potential of the substance was assessed in a weight of evidence approach including several QSAR estimations, experimental data and data on the molecular size and log Kow. The single QSAR models, studies and their results and further evidences are summarized in the table below.

EXPERIMENTAL DATA

The bioconcentration of the substance in fish was determined in a study according to OECD guideline 305 at the Institute of Ecotoxicology of the Gakushuin University (1999) with a similar substance containing O,O,O, triphenylthiophosphate as major constituent (ca. 35%).

In this study, common carp (C. carpio, 25 per group) were exposed to test substance concentrations of 0.5 and 0.05 mg/L during an 8 weeks uptake phase under flow-through conditions using vehicle. The concentrations of the major constituents were 0.188 mg/L in Level1 and 0.0188 mg/L in Level 2, respectively. Steady-state concentrations were achieved in the tissues of fish within 14 days after start of exposure. The uptake phase was followed by a 7 day depuration phase (after day 56). Test substance (constituent) recoveries throughout the test were well above 90%. 2 fish per test concentration were sampled to determine steady-state whole body BCF values for the major constituents. In addition, 2 fish per test concentration were sampled after 7 weeks exposure to determine steady-state BCF values for fish internal organs, edible parts and head-skin tissues.

The BCF values during steady-state (after 2 weeks exposure) were 1274-2508 and 1213-2194 for the constituent O,O,O, triphenylthiophosphate at the test concentrations of the test substance of 0.5 and 0.05 mg/L, respectively. Maximum values of 2508 (at day 42, level 1) and 2194 (at day 28, level 2) were recorded for the respective test concentrations. All these values are not normalised to a lipid content of 5% (fish lipid content in the study is reported to be 4.3%). No significant bioconcentration was noted for any of the other constituents (maximum BCF of 226 for all structures (< 262 normalised to 5% lipid content)). Determination of BCF values for specific fish part as well as depuration testing was performed only for O,O,O, triphenylthiophosphate. The average steady-state BCF values determined for internal organs, edible parts and head-skin tissues were 972, 438 and 793 at the 0.5 mg/L exposure level and 2918, 758 and 2511 at the 0.05 mg/L exposure level (with a maximum 4027 in internal organs). After the 7 days depuration phase significant reduction of BCF values was reported. After the 7 days depuration phase significant reduction of BCF values was reported. At the exposure level of 0.5 and 0.05 mg/L the BCF values were 146-697 and 4-15 mg/L, respectively. The half reduction time is reported as 'within 7 days'.

As for level 1 (test material concentration 0.5mg/L) swimming condition and food consumption condition and health condition were reported as unnatural (swimming), no good (food consumption), somewhat no good (health) from day one of exposure until the end of exposure and depuration the results of level one are of limited reliability. Furthermore, the solubility of the main component  of the test material (peak 1,triphenylthiophosphate ) is reported to be much lower than 0.2mg/Las applied in Level 1.

No adverse findings were reported for level 2 (test material concentration 0.05 mg/L) and the corresponding test concentration of the triphenylthiophosphate was determined to be within the solubility range (concentration of peak 1 triphenylthiophosphate ca. 0.019 mg/L). It is assumed that the adverse effects in level 1 reflect an effect due to testing above the solubilty limit rather than a toxicological effect as they do not occur in short term testing of the soluble fraction. Therefore, the BCF results in Level 2 for triphenylthiophosphate are reliable and the highest BCF value of 2194 is used as worst case.

In this study limited data on the lipid content are available. However, the fish in level 2 exposure were healthy and had good food consumption. No significant growth was observed (see attachment in robust study summary, growth data were plotted as sampling day versus natural log (ln) of weight). For further assessment the BCF value of 2194 is normalized to a lipid content of 5% resulting in a BCF of 2551. Based on the growth data it can be expected that the lipid content did not decrease to less than 2.19% which would fulfill the vB criterion (BCF > 5000) for triphenyl thiophosphate. The BCF values for the other components were far below 2000.

Therefore, the results of this study indicate that the BCF of O,O,O, triphenylthiophosphate fulfills the criterion bioaccumulative (B) but not very bioaccumulative (vB) according to REACH, Regulation (EC) No 1907/2006, Annex XIII.

QSAR

STRUCTURE A

Model/Study		BCF	logBCF	Remarks
Read-across:	Similar to OECD 305	2551		BCF values refer to main component CAS 597-82-0 at concentration of 0.0188mg/L (level 2, low test item concentration),normalized to 5% lipid content
Institute of Ecotoxicology, Gakushuin University, 1999
EPISuite v4.11	Regression-based estimate	925 L/kg	2.97	Within the applicability domain
(experimental logPow of 5.0)	Arnot-Gobas; incl. biotransformation estimates	643 - 884 L/kg	2.8 - 2.9	Within the applicability domain
	Arnot-Gobas upper trophic level; incl. biotransformation rate of zero	7882 L/kg	3.9	Within the applicability domain
US EPA T.E.S.T v.1	Consensus	215.58	2.33	Within the applicability domain of the Consensus model, confidence due to mean absolute error is low
VEGA Read across v1.0.2		136 L/kg	2.14	According to the model’s global AD index, the Read-across seems to be reliable
VEGA CAESAR v2.1.13		98 L/kg	1.99	According to the model’s global AD index, the predicted substance is out of the Applicability Domain of the model.
VEGA Meylan v1.0.2		11561 L/kg	4.06	According to the model’s global AD index, the predicted substance is out of the applicability domain of the model.
Catalogic v5.11.15		2239 L/kg (corrected)	4.12	all mitigating factors applied (main factor is size); substance is mechanistic and parametric domain but only 25% in the structural domain (75% unknown)
(experimental logPow of 5.0)		5200 L/kg (max)	4.5	No mitigating factors applied; substance is mechanistic and parametric domain but only 25% in the structural domain (75% unknown)
Other evidence Toxicological studies	No indication for bioaccumulation could be found in the available studies			(see IUCLID section 7.1)
Other evidence Log Pow 5	Screening criteria for a BCF >= 2000 are fulfilled			(see IUCLID section 4.7)
Other evidence Size (Catalogic v5.11.17)	Molecular weight and DiamMax average do not fulfill the criteria for reduced bioaccumulation accordung to ECHA guidance R.11. However, in Catalogic size (and structural flexibility) are identified as important mitigating factors.			According to the criteria in ECHA guidance on information requirements Part R.11 (2014)
DiamMax average 1.39 nm (MW 342.3)
Other evidence metabolism	indicated by Arnot-Gobas model, evidence of Cytochrom P450 induced desulfurization of organophosphates			for further details see metabolism section below

In addition to these experimentally determined BCF values a number of QSAR calculations were performed Per QSAR model the results of these calculations are tabulated and evaluated.

In general for log Kow based model calculations reliable experimental log Kow values should be preferred as model input over estimated log Kow values (e.g. KOWWIN). In case of this substance, reliable experimental data on log Kow are available .

 

The Meylan and the Arnot-Gobas predictions are considered to be reliable as these structures fall within the applicability domains of the respective sub-models. Based on the Meylan sub-model the BCF value is 925 L/kg. The BCF value calculated by the Arnot-Gobas sub-model is 7882 L/kg for the upper trophic level but a significant reduction of BCF is observed when biotransformation rates are taken into account (643 - 884 L/kg). Furthermore, a lipid content of 10.7% was assumed as default lipid content for upper trophic level in that model. Normalising to 5% lipid content would result in a BCF of 3683 L7kg. Considering biotransformation a maximum BCF value of 884 L/kg is determined. The model predictions are considered indicative for BCF values of <5000.

 

Within EPA T.E.S.T the consensus value is determined by calculating the average BCF from 5 other sub-models and is therefore considered to give the most reliable outcome. The most similar compounds displayed in the model output lack the P=S fragment of the thiophosphate which is likely to significantly change the substances’ properties and therewith can influence their bioaccumulation potential. Although the substance falls within the applicability domain of the model, the limited compound similarity is also expressed in the mean absolute errors (MAE) being partially > 0.5 compared to the training set (as well as the external test set). Therefore the confidence in the predicted BCF values is low.

 

The VEGA CAESAR sub-model calculates BCF values based on MLogP. The MLogP calculation however appears not to account for the P=S fragment in the thiophosphate moiety and is therefore likely to generate a relatively low estimated log Kow value. This however is not a prerequisite for determination of log Kow but rather a choice of molecular descriptors (i.e. fragments) used in the log Kow prediction model. The BCF value calculated with the VEGA CAESAR sub-model is 98 L/kg. The predicted MLogP value of 4.38 lies close to the experimentally determined value of 4.8-5.0. In addition to calculation of a BCF value as such, VEGA determines whether a substance falls within the applicability domain by allocating similarity indexes based on comparison of the model outcome with calculated and experimental data of structurally similar compounds from the underlying database. Based on this comparison the structure falls out of the applicability domain of the model. This is due primarily to the fact that the most similar compounds displayed in the model output contain a P=O fragment rather than the P=S fragments and therefore probably have much lower calculated and experimental BCF values than predicted for the target compounds. Although this in some cases countered by the presence of other hydrophobic moieties such as e.g. additional and/or longer alkyl chains, this still drives similarity indexes to below cut-offs for meeting the applicability domain of the model. This is visualised also in the scatter plot of experimental values for the training set form which it is clear that the target chemicals fall not within the general relationship log kow/BCF determined for the model. Based on these considerations the confidence in the predicted BCF values is considered to be low.

 

The VEGA Read-Across sub-model performs a read-across to only those 3 compounds from the dataset most similar to the target compound and determines similarity indexes based on several structural aspects of the compounds, such as their fingerprint, the number of atoms, of cycles, of heteroatoms, of halogen atoms, and of particular fragments. The estimated BCF values is calculated as the weighted average value of the experimental values of the three selected compounds, using their similarity values as weight. The BCF value calculated with this sub-model is 136 L/kg. Based on this approach the substance falls within the applicability domain of the model. However, the first two similar compounds are triphenyl phosphate (CAS 115-86-6) and triphenyl phosphite (CAS 101-02-0). These substance are known to hydrolyse faster and are readily biodegradable. Therefore the BCF prediction is considered likely to be an underestimate.

 

The VEGA Meylan sub-model is based on the Meylan model as also incorporated in the EPI Suite BCFBAF module. The log Kow values used as model input are calculated based on a slightly modified version of KOWWIN (which may explain the slightly higher log Kow values calculated in VEGA). The BCF value calculated with this sub-model is 11561 L/kg. This BCF is however calculated based on a log Kow input of 7.12 instead of the experimentally determined log Kow of 5. As the log Kow cannot be inserted manually in VEGA Meylan, and considering the Meylan log Kow/BCF relationship, this BCF should be considered to be highly overestimated. Comparable to the CAESAR sub-model, the Meylan sub-model determines whether a substance falls within the applicability domain by comparing the model outcome with calculated and experimental data of structurally similar compounds (similarity indexes). Based on this comparison the substance is just outside the applicability domain of the model (global AD Index of 0.75)

This is due to the fact that: 1. similar molecules found in the training set have experimental values that strongly disagree with the target compound predicted value; 2. reliability of logP value used by the model is not adequate.

For this substance an experimentally determined log Kow value of 4.8 -5.0 is available. This value can however not be used as input in the model (as is the case with e.g. BCFBAF). Therefore, the estimated log Kow value of 7.12 used in the VEGA Meylan BCF prediction should indeed be considered as not adequate. The Meylan BCF model is strongly based on the log Kow prediction of the compound and reaches maximum values at a log Kow of 7. The estimated BCF value based on a log Kow of 7.12 will thus be significantly overestimated. Therefore, the confidence in the predicted BCF values is considered to be low.

 

The OASIS Catalogic model BCF base-line model predicts a bioconcentration factor based on log Kow and accounts for a number of mitigating factors, such as molecular size, metabolism of parent chemical, water solubility and ionization. The experimental determined log Kow of 5 was used as model input. The highest so calculated BCF value is 5200 L/kg taking into consideration no mitigating factors. Considering the mitigating factors the BCF is 2239 L/kg. The most important mitigating factor is molecular size. However, metabolism is not regarded as significant for this structure despite the results of the Arnot-Gobas model and other indications for reduced bioaccumulation potential (see below).The structure fulfills the general properties requirements and are in the mechanistic domain of the model but are out of the interpolation structural space mainly due to structural fragments not being present in the training chemicals. Therefore, the mitigating factor metabolism might be underestimated.

 

INDICATIONS OF REDUCED BIOACCUMMULATION POTENTIAL

 

Structural alerts

The substance contains no structural fragment which is known to be related to bioaccumulation or reduced bioaccumulation, other than their contribution to estimated log Kow.

 

Log Kow

At very high Log Kow (> ca.7), a decreasing relationship with BCF is observed for organic substances. Based on current knowledge, a calculated log Kow >10 is taken as an indicator of reduced bioconcentration for PBT assessments. This cut-off does not apply to applies to O,O,O triphenyl thiophosphate

Molecular size

The decreasing log Kow/BCF relationship is considered to be due also, at least in part, by reduced uptake due to increasing molecular size. The molecular size of a substance is reflected by, among others, the average maximum diameter (Dmax-aver). Very bulky molecules will less easily pass through cell membranes which results in a reduced BCF of the substance. The substance does not fulfill the criteria for indication of reduced bioaccumulation based on molecular weight and average maximum diameter according to ECHA guidance on information requirements part R.11. However, for O,O,O triphenyl thiophosphate the OASIS Catalog model which considers size and flexibility of the structure identifies molecular size as an important mitigating factor resulting in a BCF of 2239.

 

Metabolism

The Arnot-Gobas model predicts significant biotransformation potential. In case of the Arnot-Gobas model, transformation is expected for the thiophosphate specific P=S group but also for aromatic-H and methyl groups.

However, for some organophosphorothionate pesticides cytochrome P450 mediated oxidation of the P=S group to the P=O derivative is part of the bioactivation process (defulfurization of parathion, ECETOC TR67, 1995). Furthermore, for triphenylphosphate (CAS 115-86-6), which differs only from O,O,O triphenyl thiophosphate by a central P=O fragment instead of a P=S fragment, is reported to be degraded by P-O hydrolysis in rat liver homogenate to form diphenyl phosphate as the major metabolite (EU-RAR, 2002). Although these mammalian data cannot be extrapolated directly to fish, some form of metabolism in fish is considered likely considering the rapid depuration observed in the fish BCF study.

 

Toxicological studies

No indication for bioaccumulation could be found in the available studies

 

Conclusion:

Three models give BCF values above the experimentally derived BCF of 2551: Arnot-Gobas upper trophic level; incl. biotransformation rate of zero, VEGA Meylan v1.0.2, Catalogic v5.11.15 (no mitigating factors appied). All of these modelled values have some deficiencies. In the Arnot Gobas model biotransformation was not considered and the default lipid content was assumed to be 10.7%. The BCF values in the Arnot -Gobas model range from 643 to 884 when taking biotransformation into account. In the VEGA Meylan model the experimentally derived logPow could not be used and in the Catalogic model not mitigating factors like molecular size and flexibility were not considered. The BCF modelled with Catalogic and considering mitigating factors is 2239 L/kg. On the other hand the BCF values derived with US EPA T.E.S.T v.1, VEGA Read across v1.0.2 and VEGA CAESAR v2.1.13 are likely to underestimate the bioaccumulation potential as the P=S fragment is not taken into account.

Considering the evidence for metabolism and the data from the toxicological studies limited bioaccumulation potential is expected.

In summary, in a weight-of-evidence approach balancing different QSAR estimations and experimental data significant accumulation of the compound in organisms is expected. However, due to the experimental data and the QSARs it is not expected that the BCF is above the “vB” cut-off criterion of 5000. For further assessment the BCF of 2551 is used as worst case.

STRUCTURE B - I

	Structure B	Structure C	Structure D	Structure E	Structure F	Structure G	Structure H	Structure I
Description	mono-tert-butyl O,O',O''-triphenylphosphorothioate	mono-tert-butyl O,O',O''-triphenylphosphorothioate	di-tert-butyl O,O',O''-triphenylphosphorothioate	di-tert-butyl O,O',O''-triphenylphosphorothioate	tri-tert-butyl O,O',O''-triphenylphosphorothioate	tri-tert-butyl O,O',O''-triphenylphosphorothioate	tetrakis-tert-butyl O,O',O''-triphenylphosphorothioate	tetrakis-tert-butyl O,O',O''-triphenylphosphorothioate
SMILES	CC(C)(C)c1ccccc1OP(=S)(Oc2ccccc2)Oc3ccccc3	CC(C)(C)c1ccc(cc1)OP(=S)(Oc2ccccc2)Oc3ccccc3	CC(C)(C)c1ccc(c(c1)C(C)(C)C)OP(=S)(Oc2ccccc2)Oc3ccccc3	CC(C)(C)c1ccc(cc1)OP(=S)(Oc2ccccc2)Oc3ccc(cc3)C(C)(C)C	CC(C)(C)c1ccc(cc1)OP(=S)(Oc2ccccc2)Oc3ccc(cc3C(C)(C)C)C(C)(C)C	CC(C)(C)c1ccc(cc1)OP(=S)(Oc2ccc(cc2)C(C)(C)C)Oc3ccc(cc3)C(C)(C)C	CC(C)(C)c1ccc(c(c1)C(C)(C)C)OP(=S)(Oc2ccccc2)Oc3ccc(cc3C(C)(C)C)C(C)(C)C	CC(C)(C)c1ccc(cc1)OP(=S)(Oc2ccc(cc2)C(C)(C)C)Oc3ccc(cc3C(C)(C)C)C(C)(C)C
Molecular weight	398.46	398.46	454.57	454.57	510.68	510.68	566.79	566.79
LogKow (KowWin v1.68 or measured)	8.4	8.4	10.3	10.3	12.2	12.2	14.1	14.1
Water solubility (WSKowv1.42 or measured) (mg/L)	0.0002	0.0002	0.000002297	0.000002297	0.00000002358	0.00000002358	0.0000000002391	0.0000000002391
Experimental BCF (normalised to 5% lipid content)	172-212	172-212	< LOQ	< LOQ	< LOQ	< LOQ	< LOQ	< LOQ
OASIS Catalogic v5.11.17/v5.11.15
DiamMax Average (Å)	14.531	15.718	16.127	17.347	17.831	18.519	17.685	18.780
BCFcorrected / (BCFmax) L/kg / Main mitigating factors	912 (4592) / Metabolism (0.4) / Size (0.6)	661 (4592) / Metabolism (0.3) / Size (0.7)	19 (187) / Metabolism (0.5) / Size (0.6)	15 (187) / Metabolism (0.5) / Size (0.6)	8 (16) / Metabolism (0.6) / Size (0.6) / Water solubility (0.2)	8 (16) / Metabolism (0.6) / Size (0.6) / Water solubility (0.2)	7 (9) / Metabolism (0.1059) / Size (0.099) / Water solubility (0.860)	7 (9) / Metabolism (0.1) / Size (0.1) / Water solubility (0.9)
BCFBAFv3.01
Meylan (BCF l/kg wet wt)	2820	2820	327	327	38*	38*	4*	4*
BCFBAF incl biotransformation upper, mid, low (BCF l/kg wet wt) / (biotransformation rate of zero)	490 - 772 / (2100)	490 - 772 / (2100)	20* - 31* / 45*	34* -56*/ 45*	1*-2 * / 1*	1* - 2*/ 1*	1* / 1*	1* / 1*
US EPA T.E.S.T v.1 Consensus	536	264	615	1272	1716	1120	3773	2568
VEGA Read across v1.0.2 (L/kg)	128	126	121	121	120	120	64	64
VEGA CAESAR v2.1.13 (L/kg)	56*	56*	28*	25*	15*	14*	14*	10*
VEGA Meylan v1.0.2 (L/kg)	1341	1341	155	155	18*	18*	3*	3*

* high uncertainty due to model domain

Structure B, C

Meylan predictions and the Arnot-Gobas predictionsare considered to be reliable as these structures fall within the applicability domains of the respective sub-models. Based on the Meylan submodel, the BCF is 2820 L/kg. The BCF value calculated by the Arnot-Gobas model is 2100 but a significant reduction is observed when biotransformation rates are taken into account (490 -772L/kg). Furthermore a lipid content of 10.7% was assumed as default lipid content for upper trophic level in that model. Normalising to 5% would result in an even lower BCF. Considering biotransformation a maximum BCF value of 772 L/kg is determined. The model predictions are considered to be indicative for BCF values < 2000.

Within EPA T.E.S.T the consensus value is determined by calculating the average BCF from 5 other sub-models and is therefore considered to give the most reliable outcome. The most similar compounds displayed in the model output lack the P=S fragment of the thiophosphate which is likely to significantly change the substances’ properties and therewith can influence their bioaccumulation potential. Although the substance falls within the applicability domain of the model, the limited compound similarity is also expressed in the mean absolute errors (MAE) being partially > 0.5 compared to the training set (as well as the external test set). Therefore the confidence in the predicted BCF values is low.

The CAESAR sub-model calculates BCF values based on MLogP. The MLogP calculation however appears not to account for the P=S fragment in the thiophosphate moiety and is therefore likely to generate a relatively low estimated log Kow value. This however is not a prerequisite for determination of log Kow but rather a choice of molecular descriptors (i.e. fragments) used in the log Kow prediction model. In addition to calculation of a BCF value as such, VEGA determines whether a substance falls within the applicability domain by allocating similarity indexes based on comparison of the model outcome with calculated and experimental data of structurally similar compounds from the underlying database. Based on this comparison none of the structures A to I fall within the applicability domain of the model. This is due primarily to the fact that the most similar compounds displayed in the model output contain a P=O fragment rather than the P=S fragments present in structures and therefore probably have much lower calculated and experimental BCF values than predicted for the target compounds. Although this in some cases countered by the presence of other hydrophobic moieties such as e.g. additional and/or longer alkyl chains, this still drives similarity indexes to below cut-offs for meeting the applicability domain of the model. This is visualised also in the scatter plot of experimental values for the training set form which it is clear that the target chemicals fall not within the general relationship log kow/BCF determined for the model. The BCF value for Structure B and C is < 2000 (B: 536, C: 264): Based on these considerations the confidence in the predicted BCF values is considered to be low.

The VEGA Meylan sub-model is based on the Meylan model as also incorporated in the EPI Suite BCFBAF module. The log Kow values used as model input are calculated based on a slightly modified version of KOWWIN (which may explain the slightly higher log Kow values calculated in VEGA). The BCF values calculated with this sub-model is

1341 L/kg for Structure B and C.

Comparable to the CAESAR sub-model, the Meylan sub-model determines whether a substance falls within the applicability domain by comparing the model outcome with calculated and experimental data of structurally similar compounds (similarity indexes). Based on this comparison structures B and C (BCF 56 L/kg) don't fall within the applicability domain of the model. The reasons are similar to those described above. When investigating the scatter plot of experimental values for the training set, also for the Meylan model it can be seen that the target chemicals are outliers with regard to the general relationship log kow/BCF. Although this may be partially the result of the slightly higher log Kow values calculated by the model (as compared to use of KOWWIN values in BCFBAF), for these structures the confidence in the predicted BCF values is considered to be low.

The VEGA Read-Across sub-model performs a read-across to only those 3 compounds from the dataset most similar to the target compound and determines similarity indexes based on several structural aspects of the compounds, such as their fingerprint, the number of atoms, of cycles, of heteroatoms, of halogen atoms, and of particular fragments. The estimated BCF values is calculated as the weighted average value of the experimental values of the three selected compounds, using their similarity values as weight. The BCF value for Structure B is 128 L/kg and for Structure C is 126 L/kg. Based on this approach these structures fall within the applicability domain of the model. The same considerations as presented above are however considered to apply and therefore the BCF predictions are considered likely to be underestimates.

The OASIS Catalog model BCF base-line model predicts bioconcentration factor based on log Kow and accounts for a number of mitigating factors, such as molecular size, metabolism of parent chemical, water solubility and ionization. The calculated BCF value for Structure B is 912 L/kg and for Structure C 661 l/kg taking into consideration the mitigating factors. Further, as also seen in the Arnot-Gobas model, metabolism is considered to be a significant mitigating factor. This is the case for structures B and C as according to OASIS metabolic transformation is related only to hydroxylation of the different tert-butyl groups. Another important mitigating factor is molecular size. Obviously, both mitigating factors increase with increasing number of tert-butyl groups. All structures fulfill the general properties requirements and are in the mechanistic domain of the model but are out of the interpolation structural space mainly because not all structural fragments are present in the training chemicals. Therefore, the confidence in the predicted BCF values is considered to be not high.

INDICATIONS OF REDUCED BIOACCUMMULATION

 

Structural alerts

The structures B and C contain no structural fragments which are known to be related to bioaccumulation or reduced bioaccumulation, other than their contribution to estimated log Kow.

Log Kow

At very high Log Kow (> ca.7), a decreasing relationship with BCF is observed for organic substances. Based on current knowledge, a calculated log Kow >10 is taken as an indicator of reduced bioconcentration for PBT assessments. This cut-off applies not to the structures B and Cas their calculated logKow value is 8.4.

Molecular size

The decreasing log Kow/BCF relationship is considered to be due also, at least in part, by reduced uptake due to increasing molecular size. The molecular size of a substance is reflected by, among others, the average maximum diameter (Dmax-aver). Very bulky molecules will less easily pass through cell membranes which results in a reduced BCF of the substance. For the different structures B and C the OASIS Catalog model identifies molecular size as an important mitigating factor. For the structures B and C the Dmax-aver is by OASIS calculated to be <17.4 Å which is a cut-off value at which BCF values are considered likely to be <5000 L/kg (although generally assessed in conjunction with a molecular weight of >1100 g/mol).

Metabolism

Both the Arnot-Gobas and the OASIS model predict significant biotransformation potential. In case of the Arnot-Gobas model, transformation is to expected for the thiophosphate specific P=S group but also for aromatic-H and methyl groups. According to OASIS, metabolic transformation is related only to hydroxylation of the different tert-butyl groups. No other information on metabolism is available for the structures. However, for some organophosphorothionate pesticides cytochrome P450 mediated oxidation of the P=S group to the P=O derivative is part of the bioactivation process (ECETOC TR67, 1995). Furthermore, for triphenylphosphate (CAS 115-86-6), which differs only from “Structure A” by a central P=O fragment instead of a P=S fragment, is reported to be degraded by P-O hydrolysis in rat liver homogenate to form diphenyl phosphate as the major metabolite (EU-RAR, 2002). Although these mammalian data cannot be extrapolated directly to fish, some form of metabolism in fish is considered likely considering the rapid depuration observed in the fish BCF study.

Conclusion:

Two models give BCF values above an BCF of 2000: Arnot-Gobas upper trophic level; incl. biotransformation rate of zero and Meylan predictions (BCFBAFv3.01), Catalogic v5.11.15 (no mitigating factors appied). All of these modelled values have some deficiencies. In the Arnot Gobas model biotransformation was not considered and the default lipid content was assumed to be 10.7%. The BCF values in the Arnot -Gobas model range from 490 -772 when taking biotransformation into account. In the Catalogic model not mitigating factors like molecular size and flexibility were not considered. The BCF modelled with Catalogic and considering mitigating factors is 912 and 661 L/kg. On the other hand the BCF values derived with US EPA T.E.S.T v.1, VEGA Read across v1.0.2 and VEGA CAESAR v2.1.13 are likely to underestimate the bioaccumulation potential as the P=S fragment is not taken into account.

Considering the evidence for metabolism and molecular size limited bioaccumulation potential is expected.

In summary, in a weight-of-evidence approach balancing different QSAR estimations and experimental data no significant accumulation of the compound in organisms is expected. Due to the experimental data and the QSARs it is not expected that the BCF values for both structures are above the “B” cut-off criterion of 2000.

Structure D, E

Meylan predictions are considered to be reliable as these structures fall within the applicability domain of the respective sub-model. The structures fall outside the applicability domain based on their high logKow values and are therefore more uncertain. Based on the Meylan submodel, the BCF is 327 L/kg. The BCF value calculated by the Arnot-Gobas model is 45 but a significant reduction is observed when biotransformation rates are taken into account (D: 20-31, E: 34-56 L/kg). Furthermore a lipid content of 10.7% was assumed as default lipid content for upper trophic level in that model. Normalising to 5% would result in an even lower BCF. Considering biotransformation a maximum BCF value of 31 and 56 L/kg is determined. The model predictions are considered to be indicative for BCF values < 2000.

 

Within EPA T.E.S.T the consensus value is determined by calculating the average BCF from 5 other sub-models and is therefore considered to give the most reliable outcome. The most similar compounds displayed in the model output lack the P=S fragment of the thiophosphate which is likely to significantly change the substances’ properties and therewith can influence their bioaccumulation potential. Although the substance falls within the applicability domain of the model, the limited compound similarity is also expressed in the mean absolute errors (MAE) being partially > 0.5 compared to the training set (as well as the external test set). Therefore the confidence in the predicted BCF values is low.

 

The CAESAR sub-model calculates BCF values based on MLogP. The MLogP calculation however appears not to account for the P=S fragment in the thiophosphate moiety and is therefore likely to generate a relatively low estimated log Kow value. This however is not a prerequisite for determination of log Kow but rather a choice of molecular descriptors (i.e. fragments) used in the log Kow prediction model. In addition to calculation of a BCF value as such, VEGA determines whether a substance falls within the applicability domain by allocating similarity indexes based on comparison of the model outcome with calculated and experimental data of structurally similar compounds from the underlying database. Based on this comparison none of the structures fall within the applicability domain of the model. This is due primarily to the fact that the most similar compounds displayed in the model output contain a P=O fragment rather than the P=S fragments present in structures and therefore probably have much lower calculated and experimental BCF values than predicted for the target compounds. Although this in some cases countered by the presence of other hydrophobic moieties such as e.g. additional and/or longer alkyl chains, this still drives similarity indexes to below cut-offs for meeting the applicability domain of the model. This is visualised also in the scatter plot of experimental values for the training set form which it is clear that the target chemicals fall not within the general relationship log kow/BCF determined for the model. The BCF value for Structure D and E is < 2000 (D: 28, E: 25): Based on these considerations the confidence in the predicted BCF values is considered to be low.

 

The VEGA Meylan sub-model is based on the Meylan model as also incorporated in the EPI Suite BCFBAF module. The log Kow values used as model input are calculated based on a slightly modified version of KOWWIN (which may explain the slightly higher log Kow values calculated in VEGA). The BCF values calculated with this sub-model is 155 L/kg for Structure D and E.

 

Comparable to the CAESAR sub-model, the Meylan sub-model determines whether a substance falls within the applicability domain by comparing the model outcome with calculated and experimental data of structurally similar compounds (similarity indexes). Based on this comparison structures D and E (BCF 155 L/kg) don't fall within the applicability domain of the model. The reasons are similar to those described above. When investigating the scatter plot of experimental values for the training set, also for the Meylan model it can be seen that the target chemicals are outliers with regard to the general relationship log kow/BCF. Although this may be partially the result of the slightly higher log Kow values calculated by the model (as compared to use of KOWWIN values in BCFBAF), for these structures the confidence in the predicted BCF values is considered to be low.

 

The VEGA Read-Across sub-model performs a read-across to only those 3 compounds from the dataset most similar to the target compound and determines similarity indexes based on several structural aspects of the compounds, such as their fingerprint, the number of atoms, of cycles, of heteroatoms, of halogen atoms, and of particular fragments. The estimated BCF values is calculated as the weighted average value of the experimental values of the three selected compounds, using their similarity values as weight. The BCF value for Structure D and E is 121 L/kg . Based on this approach these structures fall within the applicability domain of the model. The same considerations as presented above are however considered to apply and therefore the BCF predictions are considered likely to be underestimates.

The OASIS Catalog model BCF base-line model predicts bioconcentration factor based on log Kow and accounts for a number of mitigating factors, such as molecular size, metabolism of parent chemical, water solubility and ionization. The calculated BCF value for Structure D is 19 L/kg and for Structure E 15 L/kg taking into consideration the mitigating factors. Further, as also seen in the Arnot-Gobas model, metabolism is considered to be a significant mitigating factor. This is the case for structures D and E as according to OASIS metabolic transformation is related only to hydroxylation of the different tert-butyl groups. Another important mitigating factor is molecular size. Obviously, both mitigating factors increase with increasing number of tert-butyl groups. All structures fulfill the general properties requirements and are in the mechanistic domain of the model but are out of the interpolation structural space mainly because not all structural fragments are present in the training chemicals. Therefore, the confidence in the predicted BCF values is considered to be not high.

 

INDICATIONS OF REDUCED BIOACCUMMULATION

 

Structural alerts

The structures D and E contain no structural fragments which are known to be related to bioaccumulation or reduced bioaccumulation, other than their contribution to estimated log Kow.

 

Log Kow

At very high Log Kow (> ca.7), a decreasing relationship with BCF is observed for organic substances. Based on current knowledge, a calculated log Kow >10 is taken as an indicator of reduced bioconcentration for PBT assessments. This cut-off applies to the structures D and E as their calculated logKow value is 10.3.

 

Molecular size

The decreasing log Kow/BCF relationship is considered to be due also, at least in part, by reduced uptake due to increasing molecular size. The molecular size of a substance is reflected by, among others, the average maximum diameter (Dmax-aver). Very bulky molecules will less easily pass through cell membranes which results in a reduced BCF of the substance. For the different structures D and E the OASIS Catalogic model identifies molecular size as an important mitigating factor. For the structures D and E the Dmax-aver is by OASIS calculated to be 16.1 and 17.3 Å, which meets almost the cut-off value of 17.4 Å at which BCF values are considered likely to be <5000 L/kg (although generally assessed in conjunction with a molecular weight of >1100 g/mol).

 

Metabolism

Both the Arnot-Gobas and the OASIS model predict significant biotransformation potential. In case of the Arnot-Gobas model, transformation is to expected for the thiophosphate specific P=S group but also for aromatic-H and methyl groups. According to OASIS, metabolic transformation is related only to hydroxylation of the different tert-butyl groups. No other information on metabolism is available for the structures. However, for some organophosphorothionate pesticides cytochrome P450 mediated oxidation of the P=S group to the P=O derivative is part of the bioactivation process (ECETOC TR67, 1995). Furthermore, for triphenylphosphate (CAS 115-86-6), which differs only from “Structure A” by a central P=O fragment instead of a P=S fragment, is reported to be degraded by P-O hydrolysis in rat liver homogenate to form diphenyl phosphate as the major metabolite (EU-RAR, 2002). Although these mammalian data cannot be extrapolated directly to fish, some form of metabolism in fish is considered likely considering the rapid depuration observed in the fish BCF study.

 

Conclusion:

None of the models give BCF values above an BCF of 2000. All of these modelled values have some deficiencies. Considering the evidence for metabolism and molecular size limited bioaccumulation potential is expected.

In summary, in a weight-of-evidence approach balancing different QSAR estimations and experimental data no significant accumulation of the compound in organisms is expected. Due to the experimental data and the QSARs it is not expected that the BCF values for both structures are above the “B” cut-off criterion of 2000.

 

Structure F, G

Meylan predictions are considered to be uncertain as these structures fall out of the applicability domain based on their high logKow values and are therefore more uncertain. Based on the Meylan submodel, the BCF is 38 L/kg. The BCF value calculated by the Arnot-Gobas model is 1.4 and when biotransformation rates are taken into account in a similar range (F and G: 1.4-1.8/1.9 L/kg). Furthermore a lipid content of 10.7% was assumed as default lipid content for upper trophic level in that model. Normalising to 5% would result in an even lower BCF. Considering biotransformation a maximum BCF value of 1.8 and 1.9 L/kg is determined. The model predictions are considered to be indicative for BCF values < 2000.

 

Within EPA T.E.S.T the consensus value is determined by calculating the average BCF from 5 other sub-models and is therefore considered to give the most reliable outcome. The most similar compounds displayed in the model output lack the P=S fragment of the thiophosphate which is likely to significantly change the substances’ properties and therewith can influence their bioaccumulation potential. Although the substance falls within the applicability domain of the model, the limited compound similarity is also expressed in the mean absolute errors (MAE) being partially > 0.5 compared to the training set (as well as the external test set). Therefore the confidence in the predicted BCF values is low.

 

The CAESAR sub-model calculates BCF values based on MLogP. The MLogP calculation however appears not to account for the P=S fragment in the thiophosphate moiety and is therefore likely to generate a relatively low estimated log Kow value. This however is not a prerequisite for determination of log Kow but rather a choice of molecular descriptors (i.e. fragments) used in the log Kow prediction model. In addition to calculation of a BCF value as such, VEGA determines whether a substance falls within the applicability domain by allocating similarity indexes based on comparison of the model outcome with calculated and experimental data of structurally similar compounds from the underlying database. Based on this comparison none of the structures fall within the applicability domain of the model. This is due primarily to the fact that the most similar compounds displayed in the model output contain a P=O fragment rather than the P=S fragments present in structures and therefore probably have much lower calculated and experimental BCF values than predicted for the target compounds. Although this in some cases countered by the presence of other hydrophobic moieties such as e.g. additional and/or longer alkyl chains, this still drives similarity indexes to below cut-offs for meeting the applicability domain of the model. This is visualised also in the scatter plot of experimental values for the training set form which it is clear that the target chemicals fall not within the general relationship log kow/BCF determined for the model. The BCF value for Structure F and G is < 2000 (F: 15, G: 14): Based on these considerations the confidence in the predicted BCF values is considered to be low.

 

The VEGA Meylan sub-model is based on the Meylan model as also incorporated in the EPI Suite BCFBAF module. The log Kow values used as model input are calculated based on a slightly modified version of KOWWIN (which may explain the slightly higher log Kow values calculated in VEGA). The BCF values calculated with this sub-model is 18 L/kg for Structure F and G.

 

Comparable to the CAESAR sub-model, the Meylan sub-model determines whether a substance falls within the applicability domain by comparing the model outcome with calculated and experimental data of structurally similar compounds (similarity indexes). Based on this comparison structures F and G (BCF 18 L/kg) don't fall in the applicability domain of the model. The reasons are similar to those described above. When investigating the scatter plot of experimental values for the training set, also for the Meylan model it can be seen that the target chemicals are outliers with regard to the general relationship log kow/BCF. Although this may be partially the result of the slightly higher log Kow values calculated by the model (as compared to use of KOWWIN values in BCFBAF), for these structures the confidence in the predicted BCF values is considered to be low.

 

The VEGA Read-Across sub-model performs a read-across to only those 3 compounds from the dataset most similar to the target compound and determines similarity indexes based on several structural aspects of the compounds, such as their fingerprint, the number of atoms, of cycles, of heteroatoms, of halogen atoms, and of particular fragments. The estimated BCF values is calculated as the weighted average value of the experimental values of the three selected compounds, using their similarity values as weight. The BCF value for Structure F and G is 120 L/kg . Based on this approach these structures fall within the applicability domain of the model. The same considerations as presented above are however considered to apply and therefore the BCF predictions are considered likely to be underestimates.

The OASIS Catalog model BCF base-line model predicts bioconcentration factor based on log Kow and accounts for a number of mitigating factors, such as molecular size, metabolism of parent chemical, water solubility and ionization. The calculated BCF value for Structure F and G is 8 L/kg taking into consideration the mitigating factors. Further, as also seen in the Arnot-Gobas model, metabolism is considered to be a significant mitigating factor. This is the case for structures F and G as according to OASIS metabolic transformation is related only to hydroxylation of the different tert-butyl groups. Another important mitigating factor is molecular size. Obviously, both mitigating factors increase with increasing number of tert-butyl groups. All structures fulfill the general properties requirements and are in the mechanistic domain of the model but are out of the interpolation structural space mainly because not all structural fragments are present in the training chemicals. Therefore, the confidence in the predicted BCF values is considered to be not high.

 

INDICATIONS OF REDUCED BIOACCUMMULATION

 

Structural alerts

The structures F and G contain no structural fragments which are known to be related to bioaccumulation or reduced bioaccumulation, other than their contribution to estimated log Kow.

 

Log Kow

At very high Log Kow (> ca.7), a decreasing relationship with BCF is observed for organic substances. Based on current knowledge, a calculated log Kow >10 is taken as an indicator of reduced bioconcentration for PBT assessments. This cut-off applies to the structures F and G as their calculated logKow value is 12.2.

 

Molecular size

The decreasing log Kow/BCF relationship is considered to be due also, at least in part, by reduced uptake due to increasing molecular size. The molecular size of a substance is reflected by, among others, the average maximum diameter (Dmax-aver). Very bulky molecules will less easily pass through cell membranes which results in a reduced BCF of the substance. For the different structures F and G the OASIS Catalogic model identifies molecular size as an important mitigating factor. For the structures the Dmax-aver is by OASIS calculated to be 17.8 and 18.5 Å, which exceeds the cut-off value of 17.4 Å at which BCF values are considered likely to be <5000 L/kg (although generally assessed in conjunction with a molecular weight of >1100 g/mol).

 

Metabolism

Both the Arnot-Gobas and the OASIS model predict significant biotransformation potential. In case of the Arnot-Gobas model, transformation is to expected for the thiophosphate specific P=S group but also for aromatic-H and methyl groups. According to OASIS, metabolic transformation is related only to hydroxylation of the different tert-butyl groups. No other information on metabolism is available for the structures. However, for some organophosphorothionate pesticides cytochrome P450 mediated oxidation of the P=S group to the P=O derivative is part of the bioactivation process (ECETOC TR67, 1995). Furthermore, for triphenylphosphate (CAS 115-86-6), which differs only from “Structure A” by a central P=O fragment instead of a P=S fragment, is reported to be degraded by P-O hydrolysis in rat liver homogenate to form diphenyl phosphate as the major metabolite (EU-RAR, 2002). Although these mammalian data cannot be extrapolated directly to fish, some form of metabolism in fish is considered likely considering the rapid depuration observed in the fish BCF study.

 

Conclusion:

None of the models give BCF values above an BCF of 2000. All of these modelled values have some deficiencies. Considering the evidence for metabolism and molecular size limited bioaccumulation potential is expected.

In summary, in a weight-of-evidence approach balancing different QSAR estimations and experimental data no significant accumulation of the compound in organisms is expected. Due to the experimental data and the QSARs it is not expected that the BCF values for both structures are above the “B” cut-off criterion of 2000.

 

Structure H, I

Meylan predictions are considered to be uncertain as these structures fall out of the applicability domain based on their high logKow values and are therefore more uncertain. Based on the Meylan submodel, the BCF is 4 L/kg. The BCF value calculated by the Arnot-Gobas model is 0.9 and when biotransformation rates are taken into account in a similar range (H and G: 0.9-1 L/kg). Furthermore a lipid content of 10.7% was assumed as default lipid content for upper trophic level in that model. Normalising to 5% would result in an even lower BCF. Considering biotransformation a maximum BCF value of 1 L/kg is determined. The model predictions are considered to be indicative for BCF values < 2000.

 

Within EPA T.E.S.T the consensus value is determined by calculating the average BCF from 5 other sub-models and is therefore considered to give the most reliable outcome. The most similar compounds displayed in the model output lack the P=S fragment of the thiophosphate which is likely to significantly change the substances’ properties and therewith can influence their bioaccumulation potential. Although the substance falls within the applicability domain of the model, the limited compound similarity is also expressed in the mean absolute errors (MAE) being partially > 0.5 compared to the training set (as well as the external test set). Therefore the confidence in the predicted BCF values is low.

 

The CAESAR sub-model calculates BCF values based on MLogP. The MLogP calculation however appears not to account for the P=S fragment in the thiophosphate moiety and is therefore likely to generate a relatively low estimated log Kow value. This however is not a prerequisite for determination of log Kow but rather a choice of molecular descriptors (i.e. fragments) used in the log Kow prediction model. In addition to calculation of a BCF value as such, VEGA determines whether a substance falls within the applicability domain by allocating similarity indexes based on comparison of the model outcome with calculated and experimental data of structurally similar compounds from the underlying database. Based on this comparison none of the structures fall within the applicability domain of the model. This is due primarily to the fact that the most similar compounds displayed in the model output contain a P=O fragment rather than the P=S fragments present in structures and therefore probably have much lower calculated and experimental BCF values than predicted for the target compounds. Although this in some cases countered by the presence of other hydrophobic moieties such as e.g. additional and/or longer alkyl chains, this still drives similarity indexes to below cut-offs for meeting the applicability domain of the model. This is visualised also in the scatter plot of experimental values for the training set form which it is clear that the target chemicals fall not within the general relationship log kow/BCF determined for the model. The BCF value for Structure F and G is < 2000 (H: 14, I: 10): Based on these considerations the confidence in the predicted BCF values is considered to be low.

 

The VEGA Meylan sub-model is based on the Meylan model as also incorporated in the EPI Suite BCFBAF module. The log Kow values used as model input are calculated based on a slightly modified version of KOWWIN (which may explain the slightly higher log Kow values calculated in VEGA). The BCF values calculated with this sub-model is 3 L/kg for Structure H and I.

 

Comparable to the CAESAR sub-model, the Meylan sub-model determines whether a substance falls within the applicability domain by comparing the model outcome with calculated and experimental data of structurally similar compounds (similarity indexes). Based on this comparison structures H and I don't fall in the applicability domain of the model. The reasons are similar to those described above. When investigating the scatter plot of experimental values for the training set, also for the Meylan model it can be seen that the target chemicals are outliers with regard to the general relationship log kow/BCF. Although this may be partially the result of the slightly higher log Kow values calculated by the model (as compared to use of KOWWIN values in BCFBAF), for these structures the confidence in the predicted BCF values is considered to be low.

 

The VEGA Read-Across sub-model performs a read-across to only those 3 compounds from the dataset most similar to the target compound and determines similarity indexes based on several structural aspects of the compounds, such as their fingerprint, the number of atoms, of cycles, of heteroatoms, of halogen atoms, and of particular fragments. The estimated BCF values is calculated as the weighted average value of the experimental values of the three selected compounds, using their similarity values as weight. The BCF value for Structure H and I is 64 L/kg . Based on this approach these structures fall within the applicability domain of the model. The same considerations as presented above are however considered to apply and therefore the BCF predictions are considered likely to be underestimates.

The OASIS Catalog model BCF base-line model predicts bioconcentration factor based on log Kow and accounts for a number of mitigating factors, such as molecular size, metabolism of parent chemical, water solubility and ionization. Water solubility is considered as main mitigating factor for these structures. The calculated BCF value for Structure H and I is 7 L/kg taking into consideration the mitigating factors. All structures fulfill the general properties requirements and are in the mechanistic domain of the model but are out of the interpolation structural space mainly because not all structural fragments are present in the training chemicals. Therefore, the confidence in the predicted BCF values is considered to be not high.

 

INDICATIONS OF REDUCED BIOACCUMMULATION

 

Structural alerts

The structures H and I contain no structural fragments which are known to be related to bioaccumulation or reduced bioaccumulation, other than their contribution to estimated log Kow.

 

Log Kow

At very high Log Kow (> ca.7), a decreasing relationship with BCF is observed for organic substances. Based on current knowledge, a calculated log Kow >10 is taken as an indicator of reduced bioconcentration for PBT assessments. This cut-off applies to the structures H and I as their calculated logKow value is 14.1.

 

Molecular size

The decreasing log Kow/BCF relationship is considered to be due also, at least in part, by reduced uptake due to increasing molecular size. The molecular size of a substance is reflected by, among others, the average maximum diameter (Dmax-aver). Very bulky molecules will less easily pass through cell membranes which results in a reduced BCF of the substance. For the different structures H and I the OASIS Catalogic model identifies molecular size as an important mitigating factor. For the structures the Dmax-aver is by OASIS calculated to be 17.7 and 18.8 Å, which exceeds the cut-off value of 17.4 Å at which BCF values are considered likely to be <5000 L/kg (although generally assessed in conjunction with a molecular weight of >1100 g/mol).

 

Metabolism

Both the Arnot-Gobas and the OASIS model predict significant biotransformation potential. In case of the Arnot-Gobas model, transformation is to expected for the thiophosphate specific P=S group but also for aromatic-H and methyl groups. According to OASIS, metabolic transformation is related only to hydroxylation of the different tert-butyl groups. No other information on metabolism is available for the structures. However, for some organophosphorothionate pesticides cytochrome P450 mediated oxidation of the P=S group to the P=O derivative is part of the bioactivation process (ECETOC TR67, 1995). Furthermore, for triphenylphosphate (CAS 115-86-6), which differs only from “Structure A” by a central P=O fragment instead of a P=S fragment, is reported to be degraded by P-O hydrolysis in rat liver homogenate to form diphenyl phosphate as the major metabolite (EU-RAR, 2002). Although these mammalian data cannot be extrapolated directly to fish, some form of metabolism in fish is considered likely considering the rapid depuration observed in the fish BCF study.

 

Conclusion:

None of the models give BCF values above an BCF of 2000. All of these modelled values have some deficiencies. Considering the evidence for metabolism and molecular size limited bioaccumulation potential is expected.

In summary, in a weight-of-evidence approach balancing different QSAR estimations and experimental data no significant accumulation of the compound in organisms is expected. Due to the experimental data and the QSARs it is not expected that the BCF values for both structures are above the “B” cut-off criterion of 2000.

 

Overall comparison of the butylated phosphorothioates:

All models and the experimental results show a decreasing accumulation potential with increasing butylation except EPA TEST. The confidence in the predicted BCF decreases with butylation. Furthermore, the confidence for the predicted BCF values is low for all constituents. Considering all available data, the results of EPA tests are considered to be of low reliability and not relevant for the assessment of the bioaccumulation potential of the constituents of CAS 192268-65-8.

Other CONSTITUENTS:

The substance contains 4 minor constituents at >0.1% for which ECHA disseminated data was gathered (accessed November 8, 2017).

Based on experimental data phenol (CAS 108 -95 -2), 2 -tert-butylphenol (CAS 88 -18 -6), 4 -tert-butylphenol (CAS 98 -54 -4) and 2,4 -di-tert-butylphenol (CAS 96 -76 -4) do not fulfill the criteria for B or vB according to REACH Annex XIII and have BCF values < 2000. No significant bioaccumulation is expected.