In cleaning process validation, particularly within the matrix approach, defining worst-case products is an essential step.
It aims to identify the product (finished product, production intermediate, buffer, medium, etc.) that is most difficult to remove from equipment — the one representing the most challenging cleaning scenario, and therefore the main element to validate.
Annex 15 of the GMP guidelines introduced a concept of identification based on multi-criteria scoring to establish this choice. While it is only a recommendation, the four major criteria to consider are well known: solubility, cleanability, potency, and toxicity.
While the first three can often be assessed from physicochemical characteristics or experimental tests, toxicity remains a delicate criterion due to the diversity of available sources and data.
Classical approaches to toxicity scoring
1. HBEL values: the toxicological gold standard
When it comes to toxicity, the ideal method remains the one based on HBEL (Health-Based Exposure Limits), with reference values such as PDE (Permitted Daily Exposure) or ADE (Acceptable Daily Exposure).
These data are extremely reliable, as they are based on in-depth toxicological analyses evaluating both acute and chronic effects, and incorporating appropriate safety margins. Scientifically speaking, it is robust.
The problem is that you need to have them. And in practice, as soon as you move beyond pharmaceutical active ingredients (APIs), these values are often unavailable or incomplete.
To be clear: they could be calculated, but obtaining the full set of HBEL values for all molecules and intermediates involved in a process represents a tremendous amount of work — and above all, a significant cost. Each HBEL value must indeed be established on a case-by-case basis by a qualified toxicologist, based on a review of existing toxicological data, sometimes from the scientific literature, sometimes requiring extrapolations.
In other words, it is the most rigorous method — but also the most burdensome to implement at scale.
2. LD50
A simpler and faster approach is to turn to the LD50, which indicates the dose required to cause the death of 50% of tested animals.
This data is widely available in the literature and can serve as a first estimate of the acute toxicity of a substance.
On paper, it seems straightforward. But in reality — and for anyone who has tried to navigate this — LD50 values can be difficult to compare: oral, dermal, inhalation, rat, rabbit, mouse… So many variables that complicate interpretation and make the exercise less straightforward than it appears. In practice, LD50 can give an order of magnitude, but it remains too approximate for a consistent risk assessment.
What about OEB?
This is where a frequently underestimated but remarkably practical approach comes into play: Occupational Exposure Bands (OEB).
Occupational exposure banding is a process designed to rapidly and accurately classify chemicals into specific categories (bands), corresponding to a scale of exposure concentrations intended to protect worker health (McKernan et al. (NIOSH, 2016)).
In short, the principle is simple: each substance is classified into an exposure band (OEB 1 to OEB 5), corresponding to a range of maximum admissible concentration in air (OEL). The higher the OEB, the more toxic the product — and the more attention it deserves in worst-case selection. This data is generally accessible in Safety Data Sheets (SDS).
In essence: it is a tool designed to be fast, sound, and consistent.
Example of toxicity scoring
| Band | OEL Range | Interpretation | Score |
|---|---|---|---|
| OEB 1–2 | > 100 µg/m³ | Low to moderate toxicity | 1 |
| OEB 3 | 10 – 100 µg/m³ | Moderate toxicity | 2 |
| OEB 4 | 1 – 10 µg/m³ | High toxicity | 3 |
| OEB 5 | < 1 µg/m³ | Very high toxicity | 4 |
Why does this approach work well?
Adopting OEB in a worst-case product scoring model is a bit like having a reliable shortcut:
- Data is already available (HSE, suppliers, Safety Data Sheets…).
- The system is consistent and standardised: no need to recalculate PDE values for each molecule.
- It is universally understood: HSE professionals already know these bands, quality managers understand them, and validation teams can integrate them without difficulty.
Up to this point, we have been talking about individual molecules, whereas in real life, it is rare to work with a single pure substance. Between culture media, formulation buffers, and excipients, you quickly end up with compositions containing sometimes dozens of components.
Assigning an OEB to an isolated molecule is straightforward: one value, one band, one risk level. But for a mixture? In theory, you could calculate a global OEB for the mixture, but in practice, the necessary toxicological data are almost non-existent, particularly for culture media.
As a result, the simplest solution is to retain the most severe OEB among all substances present — in other words, the one corresponding to the most toxic molecule in the mixture. This is a rather conservative approach… sometimes a little too conservative.
Conclusion
Depending on the context, toxicity scoring can be an essential pillar in defining worst-case products. While it is ideal to rely on HBEL values (PDE or ADE) when they exist, using Occupational Exposure Bands (OEB) constitutes a robust, pragmatic, and directly usable alternative.