Registration Dossier

Environmental fate & pathways

Bioaccumulation: aquatic / sediment

Currently viewing:

Administrative data

bioaccumulation: aquatic / sediment
Type of information:
migrated information: read-across based on grouping of substances (category approach)
Adequacy of study:
key study
Study period:
From AUG to SEP 2011
2 (reliable with restrictions)
Rationale for reliability incl. deficiencies:
other: The applied model fulfils the OECD principles for QSAR-models. The submission substance is in the applicability domain of the model. Evaluation of sufficiently similar compounds with experimental data increases the confidence in the calculated BCF.

Data source

Reference Type:
study report
Report Date:

Materials and methods

Test guideline
other: REACH guidance on QSARS R.6, May 2008
Principles of method if other than guideline:
1) CAESAR QSAR model for bioconcentration factor (BCF) in fish:
Estimation methodology
Two models, Model A and Model B, have been used to build hybrid model, Model C. In the proposed approach, the outputs of the individual models (Model A and B) were used as inputs of the hybrid model. Model A was developed with a Radial Basis Function Neural Network (RBFNN) using an heuristic method to select the optimal descriptors; Model B was developed with a RBFNN using genetic algorithm for the descriptors selection. RBFNN (Wan and Harrington, 1999) was used with a Matlab function for building the models. In-house software made as a PC-Windows Excel macro was used to combine Models A and B within the Model C, using the equations defined in 4.2 of QMRF.
Input paramter: SMILES Code
For information on references and details on the model please see attached QMRF!
2) US EPA EPI SuiteTM module BCFBAF:
To support the reliable (reliability category 2) BCF from CAESAR, EPI Suite TM v4.1 (EPA, 2011) model BCFBAF (v3.01, September 2010) has been used for BCF calculation.
Input parameters: SMILES Code, log Kow (log Kow = 3.1)
Estimation methodology: Published and described in Meylan et al. (1999) for BCFWIN and updated for BCFBAF by better evaluated BCF-database as described in the program´s help file. The model is based on linear regression of log BCF against log KOW for the training set compounds. Aromatic azo compounds receive special treatment as log BCFs for compounds of log KOW between -0.02 and 9.55 were all (n=15) in the range of 0.48 and 1.82. Therefore, a value of 1.0 is assigned for all aromatic azo compounds (mean of the 15 recommended values of the training set).
3) Arnot & Gobas (2003) bioconcentration model:
To support the reliable (reliability category 2) BCF from CAESAR the Arnot & Gobas (2003) mechanistic model for assessment of the bioaccumulation potential of organic compounds in aquatic food webs was applied. The model is based on rate constants for uptake, elimination (over gills), fecal egestion (as function of dietary uptake) and growth dilution. Furthermore, the fraction of freely dissolved chemical in water is taken into account. For these constants and parameters simple relationships were developed based on default environmental parameters, assumptions on lipid content and organism weight and log Kow (sole substance specific input paramter). The model is easily transformed from bioaccumulation (including food) to bioconcentration (accumulation from the water phase only) by setting the food web biomagnification factor to zero. Calculation was performed via transformation in MS Excel spreadsheet.


Wan and Harrington, 1999. J.Chem.Inf.Comput.Sci., 39, 1049-1056.

EPA, Environmental Protection Agency (2011)
Estimation Programs Interface Suite¿ for Microsoft® Windows, v 4.10.
Online:, accessed March 2011
U.S. Environmental Protection Agency, Washington, DC, USA

Meylan, W.M.; Howard, P.H.; Boethling, R.S.; Aronson, D.; Printup, H.; Gouchie, S. (1999)
Improved method for estimating bioconcentration/bioaccumulation factor from octanol/water partition coefficient
Environmental Toxicology and Chemistry, 18, 664-672

Arnot, J.A.; Gobas, F.A.P.C. (2003)
A generic QSAR for assessing the bioaccumulation potential of organic chemicals in aquatic food webs
QSAR & Combinatorial Science, 22, 337-345

GLP compliance:

Test material


Test conditions

Details on estimation of bioconcentration:
For calculation via CAESAR BCF-model v. see attached QMRF and QPRF!
For supporting estimation via BCFBAF program v. 3.01 input parameters were SMILES code and experimental log Kow (log Kow = 3.1).
For detailed information on supporting BCF-calculation via Arnot & Gobas (2003) model see below section "Any other information on materials and methods including tables"!

Results and discussion

Bioaccumulation factoropen allclose all
9 L/kg
whole body w.w.
Calculation basis:
other: SMILES-Code
Remarks on result:
other: Key: CAESAR QSAR model for BCF in fish v.
Remarks: environment / dose:OTHER: QSAR
10 L/kg
whole body w.w.
Calculation basis:
other: SMILES-Code and log Kow
Remarks on result:
other: Supporting: BCFBAF model v. 3.01 as part of US EPA´s EPI SuiteTM v. 4.1 (EPA, 2011)
Remarks: environment / dose:OTHER: QSAR
135 L/kg
whole body w.w.
Calculation basis:
other: log Kow; 10.7% lipid content (upper trophic level organisms); default parameters
Remarks on result:
other: Supporting: Arnot & Gobas (2003) model on bioconcentration
Remarks: environment / dose:OTHER: QSAR
63.7 L/kg
whole body w.w.
Calculation basis:
other: log Kow, 5% lipid content (most experimental fish species); default parameters
Remarks on result:
other: Supporting: Arnot & Gobas (2003) model on bioconcentration
Remarks: environment / dose:OTHER: QSAR
Details on results:
For calculation via CAESAR BCF-model v. see attached QPRF!

For supporting estimation via BCFBAF program v. 3.01:
Applicability domain
Currently there is no clearly defined model domain however molecular weight and log Kow-ranges of the training set are given to be compared to the target compound. The submission substance is well in-between these limits.
The output of BCFBAF was as follows:
- Log Kow used by BCF estimates: 3.10 (user entered)
- Equation Used to Make BCF estimate:
Log BCF = 1.00 (Aromatic Azo Specification)
- Correction(s): Value
Aromatic Azo compound 0.000
- Estimated Log BCF = 1.000 (BCF = 10 L/kg wet-wt)

For supporting estimation via Arnot & Gobas (2003) model:
Applicability domain
According to Arnot and Gobas (2003) application for a wide range of organic substances is possible. Caution is required when applied for charged or ionic substances as well as surface active chemicals. For substances showing considerable dissociation information regarding uptake and bioaccumulation via the respiratory surface of aquatic organism is currently lacking.
As the submission substance is neither charged or dissociated nor a surface active chemical, it is in the applicability domain of the model.
See above section Results and discussions - Bioaccumulation factor. The results are based on a metabolic transformation rate of zero (i.e. no metabolism assumed). Calculations were performed for upper trophic level organisms with a rather high lipid content of 10.7% as well as for standard test fish species with an avarage lipid content of about 5%. Whereas the lipid content is an important parameter determining the size of the resulting BCF value temperature and organism weight are only of marginal influence on BCF.

Applicant's summary and conclusion

Validity criteria fulfilled:
To estimate BCF of the submission substance by QSAR, besides the key study (reliability category 2) performed using the CAESAR BCF model v. v. supporting calculations were performed using BCFBAF v. 3.01 and the Arnot & Gobas (2003) model on bioconcentration. The following results were achieved:
BCF according to CAESAR (key study): 9 L/kg whole body wet weight
BCF according to BCFBAF (supporting study): 10 L/kg whole body wet weight
BCF according to Arnot & Gobas model, upper trophic level (10.7% lipid content): 135 L/kg whole body wet weight
BCF according to Arnot & Gobas model, standard test fish species (5% lipid content): 63.7 L/kg whole body wet weight
Executive summary:

According to REACH Guidance R.7c section on tests according to OECD 305 states that this test might not provide a reliable BCF for compounds with low aqueous solubility (i.e. below ~10 to 100 micro-g/L). The submission substance is of extremely low water solubility. As such, a test according to OECD 305 would most certainly not result in reliable BCF data. According to the same guideline section, dietary studies are only recommended for compounds with log Kow > 6 (3.1 for the submission substance). According to R.7.10.5. describing the hierarchy of preferred data sources to determine a potential for bioaccumulation, predicted data from validated QSAR models are on rank 3 (rank 1 and rank 2: reliable BCF data from fish and invertebrates, respectively) and preferred over in vitro data. Step 3B "evaluation of non-testing data" states that "QSARs based on Kow are generally recommended if Kow is a good predictor of bioconcentration". This is the case for the application of CAESAR BCF-model on the submission substance. In accord with the here presented data the guidance states that good correlation of experimental and calculated data for analogues "increases the confidence in the BCF prediction for the substance" (see attached QPRF). According to the guidance, calculated BCF-data based on log Kow are regarded as an upper estimate which may be corroborated or reduced by strong and weak indicators as specified in table R.7.10-6. If no such indicators exist, the use of a BCF predicted from Kow is recommended (see Fig. R.7.10-1 of guidance R.7C). As the model fulfils the OECD principles for QSAR-models with algorithm and used experimental data for model build and validation being freely available; as the submission substance is in the applicability domain of the model and evaluation of sufficiently similar compounds with experimental data increase the confidence in the calculated BCF for the submission substance according to REACH guidance R.7.C; the calculated BCF value is adequate for PBT/vPvB assessment (first tier) and risk assessment (first tier) for the submission substance.

Results according to calculation via CAESAR model (key, reliability category 2; BCF = 9 L/kg ww) are in excellent agreement with the supporting results from BCFBAF program ( BCF = 10 L/kg ww).

The Arnot and Gobas BCF-model is in general conservative as it assumes no metabolism (included in regression based models via training data set) and in particular as a relatively high lipid content of upper trophic organisms is assumed. The model is independent from a training data set and the data quality thereof, as it is based on several relatively simple mechanistical assumptions relating on few environmental and organism specific parameters and the KOW of the compound in question. It is therefore completely different from regression based models dependent on the empirical database used for the regression. Furthermore as outlined by Arnot and Gobas (2003), regression based models tend to "arrive at an 'average' BCF value, allowing for a relatively large number of occurrences where the actual BCF is greater than the BCF predicted values".

Taken into account the very different methodological approach and the conservative nature of the value, the deviation by 1.2 log units of the BCF-value derived according to the Arnot-Gobas model (10.7% lipid content upper trophic level) from the log BCF of 0.93 according to CAESAR is modest and may be regarded as an upper confidence limit of the value derived by CAESAR. This is substantiated by lowering the organism lipid content to 5% (most standard test fish species) resulting in a log BCF of 1.8 (i.e. a deviation from the CAESAR derived value of 0.87 log units).