Is "BPA-Free" Actually Safe? A Guide to the Bisphenol Family

Four days ago, on 20 July 2026, the main compliance deadline passed on the EU's ban of BPA in food contact materials. Most single-use packaging sold in Europe now has to be made without it.

Which raises the question every "BPA-free" label has been quietly dodging for fifteen years: made with what instead?

The short answer is usually another bisphenol. The longer answer is that the substitutes are chemically almost the same molecule, they are about as hormonally active in a dish and in animals, one of them is considerably more potent, and the human epidemiology on them is thin enough that you can honestly read it either way. This post lays out what the family is, where you meet it, and how good the evidence actually is.

Disclosure: I'm building NeutraOat, an oral sorbent designed to capture small hydrophobic contaminants including BPA in the gut. I have an obvious interest in you concluding that this class of chemical matters. Read accordingly, and note where I say I don't have data.

What BPA actually is

Bisphenol A is two phenol rings joined by a carbon bridge. "Bis" for two, "phenol" for the rings, "A" for acetone, which is what supplies the bridge during manufacture.

Those two rings are the whole story, in both directions. Industrially, a molecule with a reactive hydroxyl group at each end is a natural building block for polymers: it links up into long chains and gives you a hard, clear, heat-resistant material. Biologically, a pair of phenol rings spaced roughly the way estradiol's are is close enough to fit an estrogen receptor. The property that makes BPA useful and the property that makes it a hormone mimic come from the same feature.

It shows up in two industrially distinct forms, and the difference matters more than most coverage admits.

Polymerized. In polycarbonate plastic and epoxy resin, BPA is chemically bound into the polymer. That covers hard clear plastics, refillable bottles, food processing equipment, and the epoxy lining inside metal food cans. Bound BPA is mostly stable, but the bonds hydrolyze slowly, and faster with heat, acidity, and age. That's why a canned tomato leaches more than a canned bean, and why an old scratched polycarbonate bottle leaches more than a new one.

Free. In thermal receipt paper, BPA isn't polymerized at all. It's loose powder in the coating, working as a colour developer that reacts with a dye when the print head heats it. A single receipt can carry milligrams of unbound BPA, which is orders of magnitude more accessible than what migrates out of a can lining, and it transfers to skin on contact. Hand sanitizer and lotion increase the absorption, because they disrupt the skin barrier right before you pick up the paper.

The rest of the family

When pressure mounted on BPA, manufacturers kept the two-phenol backbone and changed the bridge between the rings. That preserves most of the industrial behaviour, which is the point. It also preserves the part of the molecule that talks to the estrogen receptor, which is the problem.

The last two columns deliberately measure different things. The third is how hard a compound hits a receptor in a dish or a rodent. The fourth is whether anyone has shown it hurting people. A compound can score high on one and blank on the other, and for most of this family that is exactly what happens.

What the evidence actually shows

Well established

The substitutes are hormonally active. The 2015 systematic review by Rochester and Bolden pulled together the literature on BPS and BPF and found their potency in the same order of magnitude as BPA, with the same range of actions: estrogenic, antiestrogenic, androgenic and antiandrogenic, both in cells and in animals. That conclusion has held up as the literature grew.

Structural similarity predicts activity. A 2025 comparison of BPA against 26 alternatives across six in vitro assays found that several structurally close alternatives activated the estrogen receptor as strongly as BPA or more strongly. The alternatives that weren't estrogenic tended to have shifted to other modes of action rather than becoming inert. Few of the 26 came out looking better than what they replaced.

BPAF is worse, not equal. Receptor assays put it 7 to 13 times more potent than BPA. It's lower volume, so it matters less at population level, but "we replaced it with something stronger" is a fact worth sitting with.

Exposure has shifted, not fallen. Urinary BPA has declined across most monitored countries following restrictions. Over the same period BPS and BPF have risen. In US biomonitoring, BPA is detected in nearly all adults and BPS in over 90%. A 2025 cross-country analysis estimated that in Europe, roughly three quarters of bisphenol-attributable metabolic disease is now assigned to BPS or BPF rather than BPA, and in North America it splits about evenly.

The pharmacokinetic wrinkle, which cuts against the substitutes

Only unconjugated bisphenol binds the estrogen receptor. Once your liver and gut attach a glucuronide, it's inactive and on its way out. So how fast a compound gets conjugated matters as much as how potent it is at the receptor.

A 2018 pharmacokinetic modelling study measured those conjugation rates for BPA, BPS, BPF and BPAF and then modelled internal exposure. At equal external dose, BPS produced the highest internal concentration of unconjugated bisphenol of the four. The authors also found that enterohepatic recirculation, where a conjugated compound is dumped into the gut, deconjugated by bacteria and reabsorbed, appears to matter for BPS specifically.

That is the same loop that keeps PFAS in circulation for years, and it's the loop I work on, so treat my interest in this finding as suspect. But the conclusion the authors drew is the relevant one: replacing BPA with structural analogues may not reduce endocrine disruption risk, and BPS and BPAF might be more concerning than BPA once potency is accounted for.

Genuinely weaker: the human health epidemiology

Here's where I have to be honest against my own argument.

The strongest human evidence for bisphenol harm is still about BPA specifically, because BPA is what we've been measuring for twenty years. When researchers ran the equivalent analysis on the substitutes, they didn't always find the same thing. A Lancet Planetary Health analysis of NHANES data found significant associations between BPA and both general and abdominal obesity, but urinary BPF and BPS at current population exposure levels showed no significant association.

There are three readings of that, and the honest position is that we can't yet distinguish them:

1. The substitutes are genuinely less harmful in people, whatever the cell assays say.
2. Exposure is currently much lower, so the same per-molecule risk produces an effect too small to detect. The same paper noted it doesn't know whether BPF and BPS at BPA-equivalent exposure would carry the same risk.
3. It's too early. Substitution began in earnest around 2012. Chronic endocrine effects with long latency are exactly what you'd expect to miss in a cross-section fifteen years in, and BPA's own epidemiology took decades to accumulate.

Anyone claiming the substitutes are proven safe is leaning on reading one and ignoring two and three. Anyone claiming they're proven equally harmful in humans is skipping past a real null result. My own read is that reading two is most likely, because the mechanistic case is strong and the exposure gap is large and closing, but that's a judgment call and I'd rather label it as one.

Genuinely open

Most of the family has essentially no health data. There are more than twenty bisphenol analogues in commercial use and meaningful toxicology on maybe four. When a company reformulates away from BPS, the compound it moves to is usually one with less data, not more, and less data reads as safer in a regulatory system that evaluates one substance at a time.

So are the substitutes as bad as BPA?

The precise answer has three parts, and collapsing them is how both sides of this argument get it wrong.

Per molecule: yes, roughly, and BPAF is worse. Cell and animal work is consistent on this.

At current exposure levels: probably somewhat less harmful, mostly because there's less of them in people, not because they're benign. Human data is genuinely thinner and includes null results.

On trajectory: the gap is closing. BPA exposure is falling and substitute exposure is rising, and the substitutes get a longer regulatory grace period every time because they arrive with less evidence attached.

Why this keeps happening

The pattern has a name: regrettable substitution. A chemical accumulates enough evidence of harm to attract regulation, industry swaps in a close structural analogue with a similar function and a thinner file, and the evidence clock resets to zero while the exposure clock keeps running.

The mechanism is regulatory, not conspiratorial. If you assess chemicals one CAS number at a time, "we have less data on this one" functions as a defence. Building the data takes fifteen years. Reformulating takes one.

The EU's 2024 regulation is the first serious attempt to close that loop, and it's worth understanding what it did. The initial draft restricted BPA and left BPS alone, which would have produced exactly the substitution it was meant to prevent. The adopted version extended the restriction to other hazardous bisphenols, deleted both BPA and BPS from the plastics regulation, and set a default rule that other bisphenols classified as carcinogenic, mutagenic, reprotoxic or endocrine disrupting are prohibited unless specifically authorized. It also caps BPA residue in materials made with other bisphenols at 1 microgram per kilogram, which stops the substitute from smuggling the original back in.

That's class-based regulation instead of compound-based, and it's the right structural fix. The US has not done it. The FDA's stated position is that BPA is safe at the levels found in food, which puts it a long way from EFSA, whose 2023 reassessment cut the tolerable daily intake 20,000-fold to 0.2 ng/kg of body weight and concluded that dietary exposure exceeds it by two to three orders of magnitude in every age group. Germany's BfR and the EMA both objected to EFSA's methods, and BfR set its own limit a thousand times higher. This is not a field with a settled answer, and I'd be misleading you to present it as one.

What "BPA-free" tells you

It tells you one specific molecule is absent. It tells you nothing about what took its place, and the replacement is usually from the same family.

Practical version:

• Cans: BPA-free linings may be acrylic, polyester, or another bisphenol-based epoxy. A few brands use oleoresin. The label rarely says. Acidic and fatty canned foods leach the most, whatever the lining.
• Receipts: BPA-free thermal paper is usually BPS. If you want to avoid both, the term is phenol-free. Decline the receipt where you can, and don't handle one right after using hand sanitizer.
• Hard plastics: recycling code 7 is a mixed category that includes polycarbonate. Old, scratched, heat-cycled containers leach more than new ones. Glass and stainless steel sidestep the question.
• Anything heated: heat accelerates migration from every lining and every plastic. Don't microwave in plastic, however it's labelled.

The general principle is to reduce total bisphenol exposure rather than chase whichever letter is currently in the news. Chasing letters is what produced this situation.

---

Why I build for the class, not the compound

This post is really an argument for a design decision, so I'll state it plainly.

If contaminants get regulated one at a time and replaced with near-identical molecules faster than the evidence can follow, then a product that targets one named compound is obsolete on arrival. What the whole family shares is physical chemistry: small, hydrophobic, and recirculated through the gut rather than metabolized away.

NeutraOat is a modified oat fiber built to capture on those properties rather than on chemical identity. In a simulated digestive tract it captured 75% of BPA in a single pass, ahead of activated charcoal at 60%, and 99% of DEHP, a common phthalate. Those are beaker numbers, which as I keep saying is not the same as lowering anything in a person. The human pilot running right now measures PFAS, not bisphenols, so I have no human bisphenol data and won't pretend otherwise. I also have no data on BPS specifically, though the shared structure and that enterohepatic recirculation finding are the reasons I expect it to behave similarly.

First pilot results land in October 2026. I'm not going to tell you any of this works until my own data says so, and if it comes back flat I'll publish that too.

If you want to see where it goes, you can reserve a spot for a fully refundable $20 deposit, credited toward your first bottle, refunded on request, no commitment. You'll get the October pilot results before anyone else, whatever they say.

Reserve my place →

Reserving holds your place for a product in active testing. It isn't a treatment, and it isn't a drug approved to treat or prevent any disease. Nothing here is medical advice.