Toxicology Databases for Chemical Safety
Which toxicology databases should I use for chemical safety assessments?
There isn't one official list of toxicology databases you must use. The best starting point is the OECD's eChemPortal, which searches 34 government and intergovernmental chemical-information databases at once, from ECHA's REACH data to US EPA IRIS. You then add the specialist sources that match your regulatory regime and endpoint.
Approach: I read the OECD eChemPortal list of participating databases and the SCCS Notes of Guidance for cosmetic ingredients, which name specific databases and in-silico tools. Paid literature services such as PubMed/Embase-style bibliographic tools are not covered by either source.
1. Regulator datasets for substance-level hazard data (eChemPortal members)
| Purpose | Databases |
|---|---|
| EU registration study data | ECHA REACH dissemination portal [1]; OECD SIDS IUCLID [2]; OECD HPV Existing Chemicals Database [3] |
| Classification / GHS | ECHA C&L Inventory [4]; NITE Japan-GHS [5]; NZ HSNO CCID [6] |
| Reference values and risk assessments | US EPA IRIS [7]; EFSA OpenFoodTox [8]; INCHEM [9]; Canada CESAR [10]; AICIS assessments [11]; HSDB at PubChem [12]; JECDB (Japan) [13]; J-CHECK [14] |
| Pesticides / biocides | ECHA Biocides [15]; AGRITOX [16]; APVMA chemical reviews [17]; EPA Human Health Benchmarks for Pesticides [18] |
| Ecotoxicity | US EPA ECOTOX [19]; UBA ETOX [20]; EPA aquatic life benchmarks [21] |
| Computational tox / exposure and monitoring | US EPA CompTox Dashboard [22]; EU IPCHEM [23]; Nordic SPIN [24] |
| Workplace | ILO International Chemical Safety Cards [25]; German IGS [26] and ChemInfo [27] |
Two entries in the catalogue point to old sources. The UK CCRMP outputs entry links to a 2006 National Archives web capture [28]. HSDB is now reached through PubChem [12].
2. In-silico, read-across and non-animal tools (SCCS Notes of Guidance)
- General / read-across: the OECD QSAR Toolbox is described as offering "a versatile suite of programs" for categorisation, (Q)SAR and read-across. Other free tools named are VEGA, US EPA T.E.S.T. and HESS for repeated-dose toxicity. The JRC keeps an inventory of QSAR models [29].
- Genotoxicity and carcinogenicity (Appendix 10): the Danish QSAR database, OECD QSAR Toolbox, VEGA, T.E.S.T., Toxtree, Lazar, OpenTox and US EPA OncoLogic. ToxRead and VERA are named for read-across [30].
- Endocrine disruption: JRC EASIS, US EPA ToxCast, EDSP21 and ToxRefDB, US FDA EDKB and EADB, the OECD AOP knowledge base, eChemPortal, the COSMOS cosmetic-ingredient database and the Danish (Q)SAR Database [31][32].
3. Limits the SCCS points out
- Naming a tool "does not constitute an approval of its quality, or recommendation for use by the SCCS" [29].
- Regulators rely mainly on data from validated methods. In-silico results serve as supporting weight-of-evidence, not as stand-alone data [29].
- For read-across, the database "need[s] to be of high quality and sufficiently large", and analogues must be picked objectively, not by personal choice [33].
- Some of the endocrine databases "may be subject to some criticism (e.g., inaccurate information, some entries not well documented, etc.)" [31].
For a quick rule of thumb on choosing: for EU chemicals, start with ECHA REACH and the C&L Inventory, plus CLP Annex VI. For food contaminants and additives, use EFSA OpenFoodTox. For US reference doses, use IRIS. For cosmetics without animal data, use the SCCS-listed in-silico tools.
If you give me a substance (name or CAS number), I can pull its data from the REACH dossiers, its CLP classification, ATSDR profiles and minimal risk levels, EFSA OpenFoodTox, EPA ChemView and the California Prop 65 list.
Where can I find the old HSDB records?
The old HSDB records are now hosted in PubChem, which the OECD eChemPortal catalogue lists as "HSDB at PubChem" (Hazardous Substance Data Bank, https://pubchem.ncbi.nlm.nih.gov/) [12]. I can also search those legacy HSDB records directly for you, by substance name, HSDB number or PubChem compound ID.
Approach: I checked the eChemPortal entry for HSDB and searched the legacy HSDB records held in PubChem.
How the legacy records are organised
- Each record keeps its original HSDB number and is linked to a PubChem compound ID. For example, benzene is HSDB 35 and PubChem CID 241 [34]. Benzene is just an example: other HSDB titles come up the same way, such as (2-bromoethyl)benzene, HSDB 5352 / CID 7666 [35].
- The original HSDB content is still there: entries marked "PEER REVIEWED", each with its source citation. For example, the benzene record's aquatic toxicity entries include 96-hour Coho salmon data in seawater, citing Verschueren's Handbook of Environmental Data on Organic Chemicals (4th ed., 2001) and Cantelmo et al. (1982) [36].
- In PubChem, the HSDB data sits on the compound's page. The quickest route is to look up the compound by name or CAS number and open its HSDB entries.
One caveat: eChemPortal does not send searches to HSDB. It lists the database with no property data and no classification data [12], so an eChemPortal search won't show HSDB content. You need to go to PubChem itself.
If you tell me the substances you need, I can pull the HSDB entries for them, such as toxicity, environmental fate, exposure standards or medical surveillance, with citations.
Which sources help with environmental toxicity?
For environmental toxicity, use three kinds of source together. Start with the regulators' study data, mainly the ECHA REACH dossiers and US EPA ECOTOX. Then add the screening values and environmental quality standards you compare exposure against: PNECs, EPA water quality criteria and national water standards. Finally, check the hazard classification that sets labelling and the downstream obligations: CLP Annex VI aquatic classes, or a QSAR-based advisory classification where no data exist.
Approach: I checked the OECD eChemPortal catalogue for its environmental entries and opened sample records in the ECHA REACH dossiers, CLP Annex VI, the US EPA water quality criteria, legacy HSDB, ANSES AGRITOX, OFB C3PO, the University of Hertfordshire PPDB, INERIS, RIVM and the Danish EPA advisory list. I did not find EFSA OpenFoodTox's environmental endpoints or the contents of EPA ECOTOX itself in what I searched; ECOTOX appears only as a catalogue entry.
1. Study-level ecotoxicity data
| Source | What it gives you |
|---|---|
| ECHA REACH dossiers | Endpoint summaries with key values. For example, a styrene oxide algae summary gives an OECD 201 EC50 of 12 mg/L [37]. The dossiers also record waiver reasons, such as Annex IX 9.4 column 2 for terrestrial testing [38], and the registrant's PBT/vPvB conclusion [39] |
| US EPA ECOTOX Knowledgebase | Listed in eChemPortal (epa.gov/ecotox) [19] |
| UBA ETOX (Germany) | "Information System Ecotoxicology and Environmental Quality Targets" [20] |
| EnviChem | "Data Bank of Environmental Properties of Chemicals" [40] |
| Legacy HSDB (PubChem) | Peer-reviewed LC50/EC50 entries, for example Daphnia, catfish and trout values citing the EPA OPP Pesticide Ecotoxicity Database [41] |
| ECHA Biocides, AICIS, INCHEM | Biocidal active substance assessments under the BPR, Australian industrial chemical assessments and IPCS documents [15][11][9] |
eChemPortal flags ECOTOX, ETOX and the others as "Provides property data: false", so an eChemPortal property search won't return their data. You have to go to each site directly [19]. ECHA REACH is the exception: it is flagged "true" [1].
2. PNECs, benchmarks and environmental quality standards
- US EPA national recommended water quality criteria: acute (CMC) and chronic (CCC) aquatic-life values. For example, copper's saltwater CMC is 4.8 and its CCC is 3.1, and the freshwater criteria are calculated with the Biotic Ligand Model [42]. Not every substance is covered: the aquatic table had no benzene row.
- EPA OPP Aquatic Life Benchmarks for pesticides [21].
- INERIS Portail Substances Chimiques (France): curated acute and chronic ecotoxicity tables, plus regulatory PNECs (from REACH and biocides dossiers), Water Framework Directive EQS (NQE) and INERIS guide values (VGE). For benzene, it gives a 32-day Pimephales NOEC of 0.8 mg/L [43].
- RIVM Risico's van stoffen (Netherlands): Dutch environmental quality standards for water, soil and air, with their status (statutory or policy-based). It also has the Handleiding normafleiding on how to derive indicative or full environmental risk limits, which follows TGD-EQS (EC 2018), and explains how to request a standard [nl_rivm_rvs_risico_s_van_stoffen:...] (see the pages at rvs.rivm.nl/onderwerpen/normen/milieu).
3. Pesticide-specific sources
- ANSES AGRITOX: PNEC with the critical study and assessment factor. For glyphosate, the PNEC is 100 µg/L, from a chronic fish NOEC of 1,000 µg/L with an assessment factor of 10 [44].
- OFB C3PO: the same PNEC structure, compiled from open sources [45].
- PPDB (University of Hertfordshire): fate data (DT50, Koc, GUS), terrestrial, bee and aquatic endpoints, and regulatory threshold levels by species group. Each value carries a source and quality score, such as "A5 = EU regulatory data used for regulatory purposes" [46].
4. Classification
- CLP Annex VI: binding aquatic classes with M-factors. For example, copper(II) hydroxide is Aquatic Acute 1 and Aquatic Chronic 1 (H410), with acute and chronic M=10 under ATP17, in force since 17 December 2022 [47].
- Danish EPA advisory self-classification list: QSAR-predicted aquatic classes for substances that have no test data and no harmonised entry. The list itself says to follow the harmonised classification wherever one exists [48].
5. Monitoring and exposure context
EU IPCHEM provides chemical monitoring data [23], and Nordic SPIN provides use volumes in products [24].
If you give me a substance name or CAS number, I can pull its REACH ecotox summaries, PNECs, water criteria and aquatic classification in one pass.