Read-across is a predictive toxicological approach in which endpoint data from one or more source substances are used to infer the properties of a data-poor target substance based on structural similarity, mechanistic plausibility, and toxicokinetic alignment.
Within the EU regulatory landscape, read-across is embedded across multiple frameworks including:
European Chemicals Agency (REACH registrations)
European Food Safety Authority (plant protection products and biocides)
OECD test guideline waiving strategies
Under REACH, read-across must align with the Read-Across Assessment Framework (RAAF), which structures the evaluation of uncertainty via defined Assessment Elements (AEs). These AEs interrogate:
Structural similarity and category justification
Toxicodynamic similarity
Toxicokinetic comparability
Data reliability and completeness
Consistency of effects across the analogue set
However, RAAF does not quantify uncertainty. It evaluates whether uncertainty has been sufficiently addressed – not how much uncertainty remains.
Uncertainty in Read-Across is Not Optional
EFSA defines uncertainty as:
“All types of limitations in available knowledge that affect the range and probability of possible answers to an assessment question.”
In read-across, uncertainty arises from multiple domains:
Structural similarity metrics (e.g., Tanimoto indices, fragment-based similarity)
Mechanistic plausibility (AOP alignment, molecular initiating events)
Metabolic fate divergence
Differential bioavailability
Study reliability (Klimisch scoring limitations)
Endpoint-specific sensitivity to TK/TD variability
The regulatory question is not whether uncertainty exists – but whether the residual uncertainty is acceptable for the decision context.
EFSA’s 2025 Workflow – Identification Without Integration
EFSA’s updated read-across guidance (2025) explicitly identifies uncertainty at every stage of the workflow:
Problem formulation
Target substance characterisation
Source substance identification
Data matrix evaluation
Data gap filling
Overall uncertainty assessment
Yet EFSA acknowledges a critical limitation:
There is currently no harmonised framework for integrating multiple uncertainties or defining tolerable uncertainty thresholds.
It does not provide methods for semi‑quantifying uncertainty (how similar is an analogue), integrating multiple uncertainties, and what is “tolerable” uncertainty. EFSA acknowledges that there is currently "no agreed framework" for this assessment, which remains "largely based on expert opinion and interpretation”.
In practice, this leaves uncertainty integration largely dependent on expert judgement.
Moving Beyond Narrative Justification
The emerging framework proposed by Cronin & Schultz (2026) attempts to operationalise uncertainty evaluation by identifying seven discrete domains:
Chemical similarity metrics
Chemical class robustness
Toxicodynamic similarity
Toxicokinetic similarity
Common metabolite formation
Physicochemical determinants of ADME
Data quality
Each domain is scored (very low → high uncertainty), and – critically – weighted according to endpoint relevance.
For example:
Skin sensitisation → high TD impact
Repeated dose toxicity → TK + metabolite formation dominate
Developmental toxicity → mechanistic alignment is critical
This introduces a semi-quantitative dimension, shifting read-across from purely narrative argumentation toward structured transparency – allowing registrants to define "tolerable uncertainty", ensuring that the final prediction is fit for its intended purpose.
As such, their scheme represents a practical method for addressing tolerable uncertainty in a manner aligned with EFSA, OECD and ECHA expectations.
What Is Tolerable Uncertainty in Read-Across?
Regulatory acceptance hinges on fitness for purpose.
Is the read-across supporting classification only?
Or replacing a vertebrate study under Annex IX?
Or informing an EFSA risk assessment under Regulation (EC) No 1107/2009?
The higher the regulatory consequence, the lower the tolerable uncertainty.
This is where structured uncertainty mapping becomes essential – not as a defensive exercise, but as a scientific necessity.
Read-Across Uncertainty is a Feature – Not a Flaw
Read-across does not fail because uncertainty exists. It fails when uncertainty is poorly characterised.
As regulatory reliance on NAMs increases – and as animal testing continues to decline – the ability to explicitly define, contextualise, and justify tolerable uncertainty will become a core competency.
Mastering uncertainty assessment is no longer optional. It is the future of defensible read-across.
If you are considering read-across to support a REACH registration, CLP classification, EFSA submission or BPR dossier, the scientific robustness of your uncertainty assessment will determine its success.
Blue Frog Scientific specialises in designing defensible read-across strategies aligned with expectations from the European Chemicals Agency and European Food Safety Authority – ensuring uncertainty is not just acknowledged, but rigorously characterised and demonstrably tolerable.
If you'd like to discuss your read-across strategy, one of our specialist consultants with read-across expertise would be very pleased to speak with you.
