BPA Replaced by BPS, BPF and BPAF: Which One Went Where

The restrictions of the early 2010s didn't produce a substitute. They produced a family. BPA had been doing several unrelated chemical jobs at once — monomer in a clear rigid plastic, cross-linker in epoxy coatings, colour developer in till receipts — and each job needed its own replacement. Whether any of them turned out safer is a question with its own answer, and we've handled it separately: the short version is no. This is the other half of the story — which one went where, what the receptor data says about each, and what you can read off a product to tell them apart.

Every bisphenol is the same sentence with one word changed. Two phenol rings, and a bridge holding them together. BPA (CAS 80-05-7) bridges its rings with a dimethylmethylene groupA carbon atom carrying two methyl groups — the isopropylidene bridge that gives bisphenol A its name and its rigidity — one carbon carrying two methyl groups. BPS (CAS 80-09-1) uses a sulfonyl group instead, a sulphur atom with two oxygens. BPF (CAS 620-92-8) uses nothing but a bare methylene bridgeA single CH2 group — the smallest possible link between the two rings, which is why bisphenol F resins are markedly less viscous than bisphenol A resins, a single CH2. BPAF (CAS 1478-61-1) is BPA with both methyl groups replaced by trifluoromethyl groups.

That bridge is the part industry buys. It sets rigidity, thermal tolerance, viscosity, how the finished polymer behaves in a dishwasher or an oven or a printing press. The two phenol rings on the ends are the part nobody was choosing — they come along because that's what a bisphenol is. And the rings are the part your endocrine system reads.

Both ends of every one of these molecules carry a phenolic hydroxyl — the same chemical handle oestradiolThe body's principal oestrogen, and the reference compound every estrogenic-activity assay is calibrated against uses to dock into an oestrogen receptor. Changing the bridge changes the spacing between those handles, the rigidity of the whole molecule, and the electron density around the rings. It changes how well the thing docks. It does not change whether it docks.

Start with the receipt, because that market moved fastest and the numbers are cleanest. Thermal paper needs a colour developer — an acid that turns a dye black under the print head. BPA did that job for decades. When regulators restricted it, the industry mostly picked BPS, and it has stayed picked. A 2023 survey used infrared spectroscopy on 571 US thermal receipts collected in 2022 and screened by FTIR for colour developer identity receipts collected across the US in 2022 and found bisphenol S in 85% of them. Pergafast 201, the non-phenolic alternative, accounted for 12%. BPA itself was down to 1% Miller et al. 2023. Against the same team's 2017 baseline, BPA fell and BPS rose — which is exactly what a successful substitution looks like from the outside.

The hard-plastics job went to BPS too, less visibly. When polycarbonate came out of baby bottles, the polyarylsulfonesA family of high-temperature engineering thermoplastics — polyethersulfone (PES), polysulfone (PSU) and polyphenylsulfone (PPSU) — used where a plastic has to survive repeated steam sterilisation moved in: polyethersulfone and polyphenylsulfone, sold on their ability to survive sterilisation. Both are formally built from bisphenol S — PPSU pairing it with 4,4'-dihydroxybiphenyl. Migration testing on finished PPSU bottles has been reassuring: a 2018 assessment found no exceedance of specific migration limits or toxicological thresholds under milk-simulant conditions. But once those polymers weather into fragments, the picture changes — a 2022 study aged PES and PPSU microplastics derived from baby bottles under UV and measured bisphenol S coming back out, with release rising sharply with temperature Li et al. 2022. Intact bottle and weathered fragment are different exposure questions. Passing the first says nothing about the second.

BPF took the epoxy work. Its bare methylene bridge makes for a markedly lower-viscosity resin, which is why it turns up in can and tank linings, floor and pipe coatings, printing inks and dental materials. The EUEuropean Union has actually restricted its main food-contact derivative for two decades — Commission Regulation 1895/2005 prohibited bisphenol F diglycidyl ether and novolac glycidyl ethers in food contact materials from 1 January 2005, though on the procedural ground that the toxicity data never arrived rather than on a finding of harm.

And then BPF does something none of the others do. Zoller and colleagues at the Swiss federal food safety office tested 61 commercial mustards analysed for bisphenol F, of which only the mild Sinapis alba varieties contained it commercial mustards and found bisphenol F at up to 8.35 mg/kg — thousands of times the levels anyone measures migrating out of packaging. It was only in the mild ones, made from white mustard seed. The route is entirely natural: the glucosinolate glucosinalbin breaks down to 4-hydroxybenzyl alcohol, which dimerises under acid into 4,4'-BPF. Contaminated raw material and packaging migration were both ruled out Zoller et al. 2016. Liao and Kannan had already hit the same signal from the opposite direction: screening 267 US foodstuffs, they found bisphenols in about 75% of samples, and the single highest BPF concentration in the whole set — 1,130 ng/g — was a mustard Liao and Kannan 2013. A bisphenol that arrives in food with no packaging involved is the kind of result that unsettles what the word contaminant is doing.

BPAF is the odd one out, and it matters that people know it. It was never a food-contact substitute. It is a cross-linking agent for FKMFluoroelastomer — the synthetic rubber used for high-temperature, chemically aggressive seals, O-rings and gaskets fluoroelastomer seals and gaskets, and a monomer in specialty polyimides, gas-separation membranes and semiconductor processing chemicals. Consumer exposure is low enough that it is rarely quantified in biomonitoring at all. It is also, by a wide margin, the most active of the four at the receptor — which is why it took a REACH listing rather than a food-contact one. ECHA added bisphenol AF to the Candidate List of substances of very high concern on 4 February 2026, as toxic for reproduction under Article 57(c), bringing the list to 253 entries. Bisphenol S had been added three years earlier, on 17 January 2023, on two grounds at once: reprotoxicity and endocrine disruption in both humans and the environment.

The class-level answer is settled and has been for a decade. Rochester and Bolden pooled 32 studies — 25 in vitro, 7 in vivo — and concluded that BPS and BPF are as hormonally active as BPA, with endocrine-disrupting effects across estrogenic, anti-estrogenic, androgenic and anti-androgenic endpoints Rochester and Bolden 2015. A Danish group had run the same question through a battery of receptor and steroidogenesis assays the year before, testing BPB, BPE, BPF, BPS and 4-cumylphenol: same qualitative effects on estrogen and androgen receptor activity, most of the alternatives inside the same potency range as BPA Rosenmai et al. 2014. A 2024 yeast-based battery covering eleven analogues reached the same verdict and put it in the title — regrettable substitution Reininger and Oehlmann 2024.

The per-analogue detail is where it gets interesting, and BPAF is the sharpest case. Matsushima's group measured competitive binding directly. BPA displaced the reference ligand from ERα with an IC50The concentration required to displace half the bound reference ligand — lower means tighter binding of 1,030 nM; BPAF managed it at 53.4 nM. At ERβ, BPA needed 900 nM against BPAF's 18.9 nM. Roughly twentyfold and forty-eightfold tighter, respectively. But the interesting part isn't the potency — it's the direction. BPAF turned out to be a full agonist at ERα and almost completely inactive at stimulating ERβ, where it instead acted as a strong antagonist against oestradiol itself Matsushima et al. 2010. The two receptors normally pull in opposite directions on cell proliferation. Something that switches one on while blocking the other is not a stronger BPA. It's a different pharmacology.

BPS is the opposite kind of case — and the one where the standard comparison misleads. Measured at the nuclear estrogen receptor, BPS generally does come out weaker than BPA, and that is the number the substitution was argued on. Two separate findings undercut it. The first is the pathway: Viñas and Watson showed BPS triggering rapid membrane-initiated signalling in rat pituitary cells at femtomolar to nanomolar concentrations, through a route that nuclear-receptor assays don't measure Viñas and Watson 2013. The second is dose.

The dose finding is the one that changes the arithmetic. A French team gave pigs equal oral molar doses of BPA and BPS and measured what reached the bloodstream in the active, unconjugated form. BPA is famously destroyed on the way in — gut wall and liver conjugate almost all of it, leaving oral systemic bioavailability of 0.50% of an oral BPA dose reaches the bloodstream in the active unconjugated form in pigs. For BPS the figure was 57.4% of an oral BPS dose reaches the bloodstream unconjugated — roughly 115 times the fraction that survives for BPA. Plasma clearance was 3.5 times slower on top of that. Net internal exposure to the active compound came out around 250× higher for BPS than for BPA at the same dose, and the authors wrote the conclusion plainly: replacing BPA with BPS will likely lead to increased internal exposure to an endocrine-active compound Gayrard et al. 2019. Human pharmacokinetics point the same way — BPS is absorbed within an hour, cleared with a terminal half-life of about 6.8 hours, and most of the dose comes back out in urine Oh et al. 2018. Weaker per molecule, in far greater quantity, for longer, is not weaker.

The US national survey settles the exposure question. Analysing urine from 1,808 adults and 868 children in NHANESThe National Health and Nutrition Examination Survey — the US government's rolling population health and biomonitoring programme 2013–2014, BPA was detectable in 95.7% of samples, BPS in 89.4% of US urine samples contained detectable bisphenol S in NHANES 2013-2014, BPF in 66.5%. Median BPA ran higher — 1.24 µg/L in adults against 0.35 for BPF and 0.37 for BPS Lehmler et al. 2018. That gap has almost certainly narrowed since, given which way the receipt and bottle markets moved after 2014. The pattern is not local: an earlier eight-country study found BPS in 81% of 315 urine samples from the US, China, India, Japan, Korea, Kuwait, Malaysia and Vietnam, with estimated daily intakes spanning nearly two orders of magnitude between the lowest country and the highest Liao et al. 2012.

Human health-outcome evidence on the substitutes is thinner than on BPA, for the ordinary reason that they've been in wide use for less time. The clearest signal so far comes from the same NHANES data: across 3,658 adults from the 2013–2016 cycles, urinary BPS was associated with diabetes prevalence while BPF was not, with both showing associations with hypertension Moreno-Gómez-Toledano et al. 2022. Cross-sectional association, not causation, and a single analysis — it belongs in the watch column rather than the settled one. But it is the kind of result you'd expect if the endocrine-disruption mechanism transferred with the molecular shape, which is what the in-vitro work has been saying since 2014.

Read the material, not the badge. 'BPA-free' is a claim about one absent molecule; the polymer name is a chemical fact about what is present. It is usually moulded into the base, printed in the specification, or answerable by one email to the manufacturer — and unlike the badge, it can't be true and hollow at the same time.

The useful thing about tracking the substitutes one at a time is that it shows what the badge is actually measuring. 'BPA-free' records a single molecule's absence from a product. It does not record which of its structural relatives took the job, how much of that relative reaches your bloodstream instead of being destroyed on the way in, or whether the relative has been assessed at all — and on the pharmacokinetics, the relative that took most of the jobs performs worse on the measure that decides the internal dose.

The three replacements went to three different places. BPS went to the till receipt and the sterilisable bottle. BPF went to the can lining, the coating, the printing ink — and, without anybody's help, into the mustard. BPAF went to industrial seals nobody thinks about, carrying the strongest receptor activity of the four. None of that is legible from the front of a package, and all of it is legible from the material name on the back. Glass and steel never needed a badge, because they never had anything to remove.