Registration Dossier

Environmental fate & pathways

Bioaccumulation: aquatic / sediment

Currently viewing:

Administrative data

Link to relevant study record(s)

Description of key information


Key value for chemical safety assessment

BCF (aquatic species):
70.79 L/kg ww

Additional information

The aquatic bioaccumulation can be estimated with reliable methods.

The QSAR predictions and available empirical data were evaluated with respect to the following considerations:

1.     Applicability of the data to the specific information requirements for the substances in terms of relevance and adequacy.

2.     Reliability of the information according to the guidance provided in Chapter R.4 of the REACH guidance (ECHA 2008a), Annex XI of REACH and/or alternative considerations, where appropriate.

3.     Implications of the endpoint results on potential PBT conclusions based on the Integrated Testing Strategies (ITS) outlined in Chapter R.11 of the REACH guidance (ECHA 2008b).

The following QSAR models were included for bioaccumulation analysis:

·        EPIWEB-BCFWin – BCFWin is an empirical QSAR which uses a bi-linear regression through a training data set along with correction factors for certain chemical functional groups in order to predict the BCF as a function of log Kow (Meylan et al., 1999). The predicted log BCF generally increases up to log Kow of 7 and decreases thereafter. For chemicals with special behaviour (i.e., low log Kow or ionization potential), a set of rules are applied, rather than the bi-linear regression. The key input parameters for BCFWin include log Kow and chemical structure.

·        EPIWEB-Biotransformation – The Biotransformation model estimates whole body biotransformation half-lives and rate constants for organic chemicals in fish. It was developed from an evaluated data based on biotransformation estimates (Arnot et al. 2008a+b) and biotransformation is estimated based on multiple linear regressions of molecular substructures, log Kow and molecular weight. The Biotransformation model does not predict the BCF or BAF, rather the half-lives and rate constants are used as model inputs for the Arnot and Gobas BCF-BAF model.

·        EPIWEB-Arnot-Gobas BAF-BCF – The Arnot-Gobas BCF-BAF model is a mechanistic bioaccumulation model which predicts the BAF or the BCF (Arnot and Gobas, 2003). Predictions are made based on chemical log Kow, a standard set of aquatic food web and organism parameters and, if known, the chemical metabolic transformation rate (kM). The generic food web consists of 3 representative fish trophic levels with the lowest trophic level consuming prey which are assumed to be in equilibrium with the water column. The Arnot-Gobas model assumes default lipid contents of 10.7%, 6.85% and 5.98% for the upper, middle and lower trophic levels, respectively. For the assessment of the derivative oil molecules, the upper trophic level BAF and the lower trophic level BCF were considered. Note that the Arnot-Gobas model does not account for the mitigating influence of ionization on chemical absorption and may overestimate the bioaccumulation potential of ionizing substances.


The bioaccumulation predictions from EPIWEB were considered applicable for the bioaccumulation assessment because the models make predictions of either the bioconcentration factor (BCF) or the bioaccumulation factor (BAF) and they consider mitigating factors for bioaccumulation potential including ionization and metabolic transformation. The BCF estimation equation of Veith et al. (1979) which is recommended in the European Technical Guidance Document on Risk Assessment (ECB 2003) was also considered, but excluded from the assessment because it does not account for these mitigating factors and would, therefore, likely overestimate the bioaccumulation potential of the representative molecules.


The reliability of the bioaccumulation predictions was assessed according to the following requirements from Annex XI of the REACH Regulation:


·        Results are derived from a (Q)SAR model whose scientific validity has been established– Each of the bioaccumulation QSAR models have been published in peer-reviewed journal articles and the predictive capacity of each model has been evaluated using various measures of fit and in general, the models result in accurate predictions of the BCF or BAF at least 80% of the time. In addition, the models have been applied for regulatory purposes in jurisdictions outside of Europe (e.g., for Categorization of Canada’s Domestic Substances List).

·        The substance falls within the applicability domain of the (Q)SAR model– castor oil derivative falls within the molecular fragment and rule domains of the QSAR models. In addition. Instances where the substance properties were not within the model domains, and the associated implications, include:

o  The castor oil derivative is slightly higher than the MW domains for each of the models. However, this is not considered a major concern because each of the QSAR models is intended for organic chemicals and include very similar, albeit, smaller alkanoic acids in their training sets.

o  The KOWWIN-predicted log Kow values (19.2 and 21.3) fall outside of the log Kow model domains for all of the QSAR models. While this may introduce uncertainty in the predictions, the QSAR models make predictions as expected for very hydrophobic and lipophilic substances. In general, the uptake of extremely hydrophobic chemicals is expected to be very low, limiting their bionconcentration and bioaccumulation potential (for a discussion see Gobas 1993 and Gobas and Mackay, 1987). For the case of the regression-based estimate for BCF and BAF, a rule-based approach (for ionic substances with 11 or more -CH2- groups) is applied that is independent of log Kow, meaning this model domain issue does not affect the prediction. Thus, the low estimates of BAF and BCF for the KOWWIN estimates are considered appropriate for decision-making purposes, when considered within the weight ofevidence. While the predicted values may be mildly uncertain, they do demonstrate that the BCF and BAF would be far below the B threshold criteria.

Adequate and reliable documentation of the applied method is provided– the methods for the EPIWEB models are documented within EPIWEB, as well as multiple peer-reviewed journal articles and the OASIS log BCF model is also documented in a peer-reviewed journal article (see the model descriptions, above).

The results can be summarized as follows:






BCFWin log BCF



QSAR for BCF estimation (based on testing similar to OECD305)

(BCFBAF v3.00)

BCF: 70.79L/kg (whole bodyw.w. (arithmetic (anti-loggedvalue)))

(using KOWWIN logKow value16.73)


Range: 3.16 - 70.8 (See below)




Biotransformation rate constant


QSAR for biotransformation

estimation (BCFBAF v3.00)

Biotransformation rate constant (Half life): 1.6 days (whole body w.w.)

(using KOWWIN log Kow value)


Range: 1.2 - 101 (See below)




Upper trophic level log BAF (Arnot-Gobas BAF-BCF)


QSAR for BAF estimation

(BCFBAF v3.00)

BCF: 0.893 L/kg (whole body w.w. (arithmetic (anti-logged)))

 (using KOWWIN log Kow value)




Evaluation of log Kow

Each of the QSAR-predicted log Kowvalues for the molecule exceeded the screening criterion, indicating a potential for bioaccumulation. Based on this comparison, it was not possible to classify the molecule as “not B” according to the log Kowcriterion alone. As a screening criterion, the log Kowthreshold represents bioaccumulation potential based on partitioning and uptake behaviour but does not account for mitigating factors such as molecular size and biotransformation and, therefore, the BCF and BAF predictions provide a better representation of overall bioaccumulation potential. Thus, the bioaccumulation assessment relied primarily on comparison of the QSAR estimates of BCF and BAF to the definitive criteria from Annex XIII.



Using the KOWWIN estimates for log Kow, the ranges of BAF and BCF predictions are as follows:

·        Castor Oil Derivative:0.3 to 70.8L/kg;and

All of these values were far below the BCF and BAF criterion of 2000 providing support for a “not B” conclusion for the castor oil derivative.

The ranges have been calculated for all potential constituents of the substance

In conclusion, since reliable information can be obtained by means different than animal testing, this end point can be considered as covered and no further testing is needed.