In vitro skin sensitisation testing. No longer the “alternative” approach. The best science for protecting human health.
It was a real privilege to be invited to speak at the IFRA UK Sensitisers Seminar in London last week. IFRA UK does important work, and a seminar focused specifically on sensitisers brings together exactly the people whose decisions shape how safely fragrances find their way into the products we all use, from toxicologists and regulatory specialists to the safety assessors making the day-to-day calls.
I’d been asked to cover the scientific foundations of Type IV skin sensitisation and how to assess fragrances. That’s familiar ground for me, and a genuine pleasure to walk through with people who care about the detail. But I wanted to spend some of my time on something I find myself returning to in conversations across the industry. It’s the word “alternative,” and how I think we’ve outgrown it.
Let me explain why with a number.
When you’re assessing whether a fragrance ingredient is going to cause allergic contact dermatitis, one of the tests that’s been treated for many years as a reference point is the Local Lymph Node Assay, or LLNA. It’s a mouse test. Here’s the part I’d like every safety assessor to know, because it’s quietly remarkable: when the LLNA is compared against actual human reference data, its balanced accuracy is 57.9%. That figure isn’t ours. It sits in OECD Guideline 497, Table 1.2, in black and white. It’s genuinely little better than tossing a coin, so why do it?
I could see the surprise on a few faces when that number went up on the slide. Honestly, I’d be more worried if it didn’t get a reaction. For a test that’s shaped how this industry has approached skin sensitisation for years, 57.9% against human data is the kind of figure that quite reasonably stops a room. Someone afterwards said to me they’d been working with the LLNA for most of their career and had never seen it laid out quite like that.
We also took a look at what the in vitro and in chemico methods, the ones now codified in that same guideline under the “2 out of 3” Defined Approach, are achieving. The strongest combinations are coming in at 97% balanced accuracy against human data. The pattern across the full table is consistent and it’s the good kind of consistent. Combinations of human-relevant tests are sitting well above what the mouse model managed, often by 30 percentage points or more.
This is wonderful news. The methods we have today, built on human cells, human proteins and human-relevant chemistry, are giving us better answers than we’ve ever had. And that’s where my point about language comes in.
We still call these methods “alternative.” But alternative implies there’s a proper version somewhere, and this is the substitute. It carries the quiet suggestion of second best. The science has now moved beyond that. These methods aren’t an alternative to good science. They are the most accurate tools we have for predicting how human skin will respond to a substance. The relationship between the methods and the word has, gently and without anyone meaning it to, ended up the wrong way around.
The regulators are starting to use a better term, New Approach Methodologies, or NAMs. I’d love to see that one take hold. It captures what these methods actually are, which is the current state of the art for predicting human outcomes from human biology.
What was lovely about the conversations on the day was how much momentum there is behind this shift. The UK Government’s “Replacing Animals in Science” strategy, launched in November 2025, identifies skin sensitisation as one of the endpoints where animal testing will be phased out from the end of 2026. The Defined Approaches in GD497 already meet, and in most permutations exceed, what they’re replacing. The science is finally in a position where it can confidently say: yes, we can do this, and we can do it better.
There’s one more thread I always come back to, because it’s the reason I founded XCellR8. Even within in vitro testing, “animal-free” doesn’t always mean what people might assume. Most cell-based methods still use foetal bovine serum, animal-derived antibodies or tissue extracts somewhere in the process. That’s why we built the XCellR8 scale to make those distinctions visible, and why we’ve spent years adapting methods like KeratinoSens™ to genuinely animal-product-free conditions (an adaptation that’s now annexed in the OECD Test Guideline itself). For brands making cruelty-free or vegan claims, the supply chain behind the safety data matters as much as the headline method.
So I left London feeling positive, as I usually do after these events. The science has moved. The regulation is moving with it. The numbers in GD497 are doing more to make the case for human-relevant testing than any of us could from a stage. The last piece, and the easiest piece really, is the language. The methods we have aren’t alternatives. They are simply the best tools we have for the job.
If you’re working through any of this for your own ingredients, do get in touch. It’s exactly the kind of conversation we love having.
———
Contact us if you’d like to learn how you can support our mission, become a corporate sponsor or donate to our crowdfunder.
Email info@x-cellr8.com
Call +44 (0)1925 607 134.
Follow Us!