Five NSAIDs from one pharmacological class have five different transdermal fates. No molecular descriptor predicts which — and neither does the flux calculation that comes closest.
Transdermal development is usually taught as a screening problem: check the molecular weight, check the log P, and the candidate either passes or it does not. The marketed products tell a harder story. Every NSAID that reached the market as a patch clears the screen — and so does the one that never did. The molecule with the highest predicted flux has the smallest franchise, and the two with the lowest have the largest.
A patch is not a dosage form
A tablet’s formulation is working to get the drug through the gastrointestinal tract. A patch’s formulation is working to persuade the body’s largest organ to let the drug in — and the stratum corneum, corneocytes set in a lipid matrix, is extremely good at refusing. The useful question is therefore not “can this drug be put into a patch?” but whether the molecule, the vehicle, the dose, the route and the clinical objective can be made to agree at once. A patch is not a dosage form. It is a drug–vehicle–skin–device system, and the molecule is one of five terms.
What the 500-Dalton rule rests on
The field’s foundational screen is the 500-Dalton limit, set out by Bos and Meinardi in Experimental Dermatology in 2000: “the molecular weight of a compound must be under 500 Dalton to allow skin absorption. Larger molecules cannot pass the corneal layer.”
The paper gives three arguments. Virtually all common contact allergens are under 500 Da. The most commonly used agents in topical dermatotherapy are under 500 Da. And, verbatim: “all known topical drugs used in transdermal drug-delivery systems are under 500 Dalton.”
Only the third is circular — an observation about the installed base offered as evidence for the constraint that produced it. The first is not: contact allergens are thousands of industrial and cosmetic chemicals never selected for delivery success. And the size term has mechanistic grounding that owes nothing to marketed products; Potts and Guy fitted it in 1992 to some ninety largely non-pharmaceutical compounds, giving log kp = 0.71 log P − 0.0061 MW − 2.72.
The rule is sound, then, as a necessary condition — but never designed to discriminate among molecules that all clear it. Roughly two dozen actives have ever reached the market as patches, at about one approval every two years, and rotigotine is the only new chemical entity ever developed specifically for the route.
Five NSAIDs, five fates
The NSAIDs make the point better than any abstraction, because pharmacological class is held constant and the outcomes still diverge.
Diclofenac is the only NSAID with an FDA-approved topical dermal product; ophthalmic and nasal NSAIDs aside, a sweep of Drugs@FDA returns diclofenac and nothing else. Flector (10 × 14 cm, 180 mg diclofenac epolamine, twice daily) and Licart (182 mg, once daily) are marketed in the US, and diclofenac tapes in Japan. The UK had one too — Voltarol 140 mg medicated plaster, PL 15545/0012 — and its SmPC now carries the status Discontinued, which is its own comment on how contingent these markets are.
Ibuprofen has a 200 mg medicated plaster authorised nationally in the UK (PL 00063/0733, first authorised 9 June 2016) and in a number of EU member states — Nurofen Joint & Muscular Pain Relief in Britain, Nurofen Durance in Ireland, Nurofen 24-Stunden Schmerzpflaster in Germany. It has nothing in the US.
Naproxen has no patch or plaster in any of those registers. (Topical naproxen gels are marketed in some European territories; the patch is what is missing.)
And then there is Japan, which the Western literature routinely leaves out of this comparison and which is the largest medicated-plaster market in the world. Alongside its diclofenac tapes it markets ketoprofen (Hisamitsu’s Mohrus line, the biggest NSAID plaster franchise of any), loxoprofen, felbinac, flurbiprofen and indomethacin as tapes and poultices. Any account of “which NSAIDs became patches” that stops at two continents is describing a fragment.
One of those gaps is a paperwork gap. The US OTC monograph for external analgesics, M017, permits none of the arylalkanoic-acid NSAIDs — no diclofenac, ibuprofen, naproxen or ketoprofen — and both its active-ingredient sections open with the same exclusion: the listed actives are permitted “but not for use in a patch, plaster, or poultice dosage form.” Methyl salicylate is admitted at 10–60%, but only as a counterirritant and only outside a plaster. There is no monograph route to any OTC medicated plaster in the United States; topical diclofenac gets there on an NDA instead. The Atlantic asymmetry is regulatory architecture, not a permeability finding.
The descriptor that cannot separate them
Naproxen is 230 Da at log P 3.18, ketoprofen 254 at 3.12, diclofenac 296 at 4.51, ibuprofen 206 at 3.97, indomethacin 358 at 4.27. On the two descriptors the screening matrix actually asks for, all five sit inside the window and nothing separates them.
So compute the quantity that is supposed to govern delivery instead — maximum flux Jmax = kp × Csat, permeability coefficient times saturated concentration, with kp from the Potts–Guy regression. It does not rescue the picture. It produces a ranking that the market contradicts.
| Molecule | MW (Da) | log P | Predicted Jmax (µg/cm²/h) | Reached the market as |
|---|---|---|---|---|
| Ibuprofen | 206 | 3.97 | 1.45 | UK/EU plaster only |
| Ketoprofen | 254 | 3.12 | 0.45 | Japan — the largest franchise |
| Naproxen | 230 | 3.18 | 0.22 | no patch anywhere |
| Diclofenac | 296 | 4.51 | 0.11 | US · Japan (UK plaster discontinued) |
| Indomethacin | 358 | 4.27 | 0.013 | Japan |
Ibuprofen has the highest predicted flux of the five and the narrowest market. Ketoprofen, at under a third of ibuprofen’s, has the largest plaster franchise in the world. Indomethacin, two orders of magnitude below ibuprofen, is marketed as a tape. Diclofenac, the only NSAID with an FDA-approved dermal product, sits fourth of five.
Three caveats belong with those numbers, and they matter more than the numbers do. The Potts–Guy regression was fitted to nonelectrolytes; all five of these molecules are carboxylic acids with pKa around 4.0–4.4, so the model is being used outside the domain it was built for. Its residual scatter is roughly 0.7–0.8 log units, which means the spread across these five is not much larger than the method’s own error. And no marketed product contains the species the calculation describes — Flector and Licart carry diclofenac epolamine, Voltarol and the Japanese tapes diclofenac sodium. Salt selection changes how much drug a vehicle can carry and how fast it dissolves; it changes neither skin’s permeability coefficient nor, at equilibrium, the maximum flux. Its value is kinetic and formulative — supersaturation and adhesive loading — not a higher thermodynamic ceiling.
So what does a candidate actually need?
Log P earns its place in the matrix, but not for the reason usually given. A permeant has to do two opposite things: partition into the lipid stratum corneum, then partition out into the aqueous viable epidermis beneath it. Lipophilicity helps the first step and hinders the second, which is why there is an optimum rather than a trend — and why a very lipophilic molecule can enter the barrier readily and then sit in it.
There is a second, sharper reason. Log P appears twice in Jmax = kp × Csat, with opposite signs: it raises the permeability coefficient (Potts–Guy coefficient +0.71) and lowers aqueous solubility (roughly −1.0 by the general solubility equation). The effects do not merely cancel — they over-cancel, leaving a net dependence of maximum flux on lipophilicity of about −0.3. Magnusson and colleagues showed the consequence directly in the Journal of Investigative Dermatology in 2004: permeability coefficients correlate best with log P, but maximum flux is governed by molecular weight, log Jmax = −3.90 − 0.0190 MW, with molecular weight alone accounting for 85% of the variance across 87 solutes. Melting point and hydrogen-bond acceptor capability each added more to that regression than lipophilicity did.
The same paper contains the warning that should travel with it. Validated against ionisable solutes — the class every NSAID belongs to — the same equation explains 28% of the variance, not 85%. These models sort molecules into broad bands. They do not adjudicate between neighbours.
So the honest parameter list looks like this:
| Parameter | Wanted | Why | What fails |
|---|---|---|---|
| Molecular weight | < 500 Da, and smaller is strongly better | Diffusion through lipid lamellae is free-volume limited, so the size penalty is exponential, not linear | Above ~500 Da flux collapses; no marketed passive patch carries an active much above 470 Da |
| Melting point | < 200 °C | A high-melting solid has a strong crystal lattice, which depresses solubility in every solvent | Nothing dissolves it enough to reach a useful concentration in any vehicle |
| Saturated concentration | as high as the vehicle will hold | Flux is driven by thermodynamic activity; Csat scales Jmax directly | A permeable molecule at low Csat still delivers nothing |
| Log P | roughly 1–4; an optimum, not a threshold | Must partition into the stratum corneum and then out of it — two opposing requirements | Below ~1 it never enters the lipid phase; well above ~4 it enters and stays, and solubility collapses. Diclofenac at 4.51 sits outside the usual window and is still the only NSAID with an FDA-approved dermal product |
| Hydrogen bonding | few donors and acceptors | Each hydrogen bond must be broken to enter the lipid domain | Depresses permeability independently of size and lipophilicity |
| Ionisation | the neutral species is the one that permeates | Only the un-ionised fraction partitions appreciably into lipid; with excess solid present, maximum flux is set by the intrinsic solubility of that neutral species and is largely pH-independent | Reading a salt’s higher solubility as higher flux. Vehicle pH is the lever the formulator holds; skin pH is a boundary condition |
| Daily dose | ≤10–20 mg/day at conventional areas | Required area = dose ÷ (24 × flux), so dose alone means nothing without the other two | Hundreds of milligrams is not wearable at any demonstrated flux. Nicotine at 21 mg/24 h and methylphenidate at 30 mg/9 h show the ceiling is higher than the folklore |
| Skin tolerability | low irritation and sensitisation | Twenty-four hours of occlusion under an adhesive is a far harsher exposure than a rub-on gel | A tolerable gel can still be an intolerable patch |
Only two of those eight — molecular weight and log P — are what the standard screen asks for. The other six are where the answer lives, and the last one is not a molecular property at all.
What the arithmetic rules out
Wiedersberg and Guy reduce feasibility to a single relation between daily dose, achievable flux and area:
(1000 / 24) × Q = A × Jmax, where Q is the daily dose in mg, A the patch area in cm², and Jmax the flux in µg/cm²/h.
Before running it, one correction to how the marketed products are usually read. They are not a map of what skin permits. Wiedersberg and Guy tabulate each patch’s real in vivo flux against its predicted passive maximum, and the ratios run both ways by more than an order of magnitude: Neupro’s rotigotine delivers 26 times its predicted maximum and Daytrana’s methylphenidate 29 times, on enhancer chemistry and the thermodynamic state of the drug in the adhesive. Nicotine runs the other way — its predicted maximum is about 1,425 µg/cm²/h and the marketed systems deliver at 0.02 to 0.05 of that, because, as the authors put it, nicotine “permeates the skin very quickly, to an extent which is far greater than necessary.” Nicotine’s 29 µg/cm²/h is a throttle setting, not a ceiling.
Now run it on a systemic naproxen patch replacing an oral 500 mg/day regimen — the lowest end of the UK maintenance range, which runs to 1 g daily. At naproxen’s own predicted maximum of 0.22 µg/cm²/h the patch needs 96,000 cm², about 9.6 square metres. Grant it the largest vehicle multiple any marketed patch has demonstrated — Daytrana’s twenty-nine-fold — and it still needs 3,330 cm², more than five A4 sheets. The same calculation puts ibuprofen at 1,190 cm² for 1,200 mg/day.
Note which way that runs. The molecule with the worse dose arithmetic is the one with the plaster — the tell that the plasters were never solving this problem. Every NSAID plaster marketed in the US and Europe is indicated for local treatment; the ibuprofen plaster’s SmPC reports a mean day-5 Cmax of 0.51 µg/mL against a stated 20–50 µg/mL for a 200–400 mg oral dose.
Except that a marketed product says the 29-fold ceiling is too low. On 23 March 2021 Japan’s Ministry of Health, Labour and Welfare approved Zicthoru, a Hisamitsu tape carrying 75 mg of diclofenac sodium in 70 cm², two applied once daily and increasing to three at maximum in the cancer-pain indication; the indications were extended in June 2022 beyond cancer pain to low back pain and three further musculoskeletal conditions, for which the maximum is two tapes. The package insert caps the dose with an explicitly systemic instruction: because exposure with three tapes reaches approximately the same level as that with the usual dose of oral diclofenac sodium (author’s translation of the Japanese package insert), no more than three may be worn. That three-tape maximum, and the systemic caution attached to it, belong to the cancer-pain indication.
Work backwards from that. Three tapes are 210 cm². Delivering tens of milligrams a day across that area implies an in vivo flux somewhere around 8–10 µg/cm²/h — seventy to ninety times the model’s prediction for diclofenac. Grant naproxen and ibuprofen a comparable multiple and their systemic patches come out at roughly 1,100 cm² and 400 cm². Large. Not nine square metres.
Which is the finding, and it is not the one the calculation was set up to produce. The model’s spread across five molecules is about 130-fold. Its error against a single real product is around 80-fold. A model that misses a marketed patch by two orders of magnitude cannot be what decided which molecules got one. What Zicthoru actually demonstrates is that the fifty-square-centimetre convention was a convention — a habit of the installed base, not a limit anyone had established — and that the vehicle and the wear time, the two terms the screen never asks about, are where the delivered dose came from.
Potency still matters, and it is the property most often left off the list: at any given flux the required area scales with the daily dose, which is why diclofenac at roughly a sixth of naproxen’s oral dose is the NSAID with a marketed systemic transdermal product in Japan, and naproxen is not. But the flux model does not predict that outcome. It ranks diclofenac fourth of five.
Was the local route worth it anyway
The question the screen never asks is whether skin delivery buys anything. For local NSAIDs the answer is now guideline-level: NICE guideline NG226 (England, 19 October 2022) tells clinicians to offer a topical NSAID for knee osteoarthritis (1.4.2) and to consider one for other affected joints (1.4.3), placing oral NSAIDs behind them — “if topical medicines are ineffective or unsuitable” (1.4.4) — and to offer a gastroprotective treatment to people with osteoarthritis while they are taking an NSAID.
The trade is quantified. Andersen and colleagues, in Clinical Pharmacology in Drug Development in 2025, ran two crossover studies against oral diclofenac 50 mg three times daily: day-7 AUC from diclofenac diethylamine gel was some forty-two-fold lower than the tablets and peak concentration about three-hundred-fold lower. These were pharmacokinetic studies. Lower systemic exposure is a measured pharmacokinetic difference; neither study measured gastrointestinal or cardiovascular events, and no comparative safety conclusion follows from them here.
And the drug arrives. Seefried and colleagues sampled synovial tissue and fluid from forty-five patients awaiting knee arthroplasty and found measurable diclofenac in every treated participant twelve hours after the last dose. Note what that study does not show, since it is routinely miscited: tissue:plasma and fluid:plasma ratios were 0.32 and 0.46 — higher in plasma than in the joint at trough. The defensible claim is penetration and persistence, not preferential accumulation.
Where these programmes actually die
Neither of the two most instructive recent failures died on permeation.
Zosano’s Qtrypta, a zolmitriptan microneedle system, drew an FDA Complete Response Letter in October 2020 citing “inconsistent zolmitriptan exposure levels” and “inadequate pharmacokinetic bridging” between manufacturing lots. The drug crossed. It crossed differently on Tuesday than it had on Monday. FDA issued a Refuse to File letter on 17 February 2022; the company filed for Chapter 11 that June.
Radius Health’s abaloparatide transdermal system is the sharper lesson, because it worked. In Phase 3 it raised lumbar spine bone mineral density by 7.1% at twelve months against 10.9% for the subcutaneous injection — 3.7 percentage points against a pre-specified non-inferiority margin of 2.0%, missing by 1.7. FDA indicated another pivotal trial would be needed; Radius put that at “$100+ million of additional capital … over the next 3+ years” and stopped in June 2022.
A drug was delivered across human skin, at scale, in a registrational trial, and the programme still ended — because it delivered less than the needle it replaced. Crossing the barrier was never the hard part. Crossing it reproducibly, and enough, was.
The molecule must earn it
The screening matrix that opens most transdermal programmes asks whether the molecule can cross — the least discriminating question available. All five NSAIDs pass, and their fates are entirely different. What separated them was potency, the route the product was aimed at, what the vehicle and the wear time could be made to deliver, the area a person would wear, and the regulatory architecture of the market — in one case, whether a monograph would permit a plaster at all.
The 500-Dalton rule survives all of that as a necessary condition. What it cannot do is choose. Neither can the flux model built on top of it, which ranks the market’s biggest NSAID plaster franchise fourth and misses a marketed systemic patch by two orders of magnitude. The screen is cheap; the conclusion drawn from it is not.
A patch is not a dosage form. The molecule does not arrive with the right to become one. It has to earn it.
Sources
- Bos JD, Meinardi MMHM. The 500 Dalton rule for the skin penetration of chemical compounds and drugs. Exp Dermatol 2000;9(3):165–169. https://pubmed.ncbi.nlm.nih.gov/10839713/
- Potts RO, Guy RH. Predicting skin permeability. Pharm Res 1992;9(5):663–669. https://pubmed.ncbi.nlm.nih.gov/1608900/
- Potts RO, Guy RH. A predictive algorithm for skin permeability: the effects of molecular size and hydrogen bond activity. Pharm Res 1995;12(11):1628–1633. https://pubmed.ncbi.nlm.nih.gov/8592661/
- Magnusson BM, Anissimov YG, Cross SE, Roberts MS. Molecular size as the main determinant of solute maximum flux across the skin. J Invest Dermatol 2004;122(4):993–999. https://doi.org/10.1111/j.0022-202X.2004.22413.x
- Wiedersberg S, Guy RH. Transdermal Drug Delivery: 30+ Years of War and Still Fighting! J Control Release 2014;190:150–156. https://purehost.bath.ac.uk/ws/files/62445356/Accepted_version.pdf
- Pastore MN, Kalia YN, Horstmann M, Roberts MS. Transdermal patches: history, development and pharmacology. Br J Pharmacol 2015;172(9):2179–2209. https://pmc.ncbi.nlm.nih.gov/articles/PMC4403087/
- FLECTOR (diclofenac epolamine) topical system 1.3% — US prescribing information. https://dailymed.nlm.nih.gov/dailymed/drugInfo.cfm?setid=1f19a016-c4fc-43c5-9990-e5f2e2776a82
- LICART (diclofenac epolamine) topical system 1.3% — US prescribing information. https://dailymed.nlm.nih.gov/dailymed/drugInfo.cfm?setid=851c8b39-6056-4bbc-812e-b9b13977c1ac
- Nurofen Joint & Muscular Pain Relief 200mg Medicated Plaster — SmPC, PL 00063/0733. https://www.medicines.org.uk/emc/product/9321/smpc
- Voltarol 140mg Medicated Plaster — SmPC, PL 15545/0012. https://www.medicines.org.uk/emc/product/6992/smpc
- FDA OTC Monograph M017, External Analgesic Drug Products for Over-the-Counter Human Use. https://www.accessdata.fda.gov/drugsatfda_docs/omuf/monographs/OTC%20Monograph_M017-External%20Analgesic%20Drug%20Products%20for%20OTC%20Human%20Use%2005.02.2023.pdf
- Naprosyn 500mg Tablets — SmPC, PL 43252/0005. https://www.medicines.org.uk/emc/product/13238/smpc
- ZICTHORU Tape 75 mg — PMDA electronic package insert and Hisamitsu interview form. https://www.hisamitsu-pharm.jp/product/data/zicthoru_i.pdf
- DAYTRANA (methylphenidate transdermal system) — US prescribing information. https://www.accessdata.fda.gov/drugsatfda_docs/label/2017/021514s028lbl.pdf
- NICE guideline NG226. Osteoarthritis in over 16s: diagnosis and management, 19 October 2022 (recommendations 1.4.2–1.4.4). https://www.nice.org.uk/guidance/ng226
- Andersen G, Bariguian Revel F, Armogida M. Two Phase I, randomized, open-label, multiple-dose, crossover studies investigating the systemic bioavailability of topical diclofenac diethylamine gels compared with oral diclofenac tablets. Clin Pharmacol Drug Dev 2025;14(9):688–699. https://doi.org/10.1002/cpdd.1560
- Seefried L, Blyth M, Maheshwari R, et al. Penetration of topical diclofenac into synovial tissue and fluid of osteoarthritic knees. Ther Adv Musculoskelet Dis 2020;12:1–13. https://doi.org/10.1177/1759720X20943088
- Zosano Pharma receives Complete Response Letter from FDA for Qtrypta, 21 October 2020. https://www.globenewswire.com/news-release/2020/10/21/2111872/0/en/Zosano-Pharma-Receives-Complete-Response-Letter-from-FDA-for-Qtrypta.html
- Zosano Pharma receives Refuse to File letter from FDA for Qtrypta, 17 February 2022. https://www.globenewswire.com/news-release/2022/02/17/2387192/0/en/Zosano-Pharma-Receives-Refuse-to-File-Letter-from-FDA-for-Qtrypta.html
- Radius Health announces results from the wearABLe trial, 8 December 2021. https://www.globenewswire.com/news-release/2021/12/08/2348411/31149/en/Radius-Announces-Results-from-the-wearABLe-Trial-Evaluating-Abaloparatide-Transdermal-System-in-Postmenopausal-Women-with-Osteoporosis.html
- Radius Health update on abaloparatide transdermal system, 1 June 2022. https://www.globenewswire.com/en/news-release/2022/06/01/2454734/31149/en/Radius-Health-Update-on-Abaloparatide-Transdermal-System.html
- Hisamitsu Pharmaceutical — MOHRUS (ketoprofen) transdermal product family. https://global.hisamitsu/products/
- PubChem, National Library of Medicine — molecular weight, experimental log P and water solubility for ibuprofen (CID 3672), naproxen (CID 156391), ketoprofen (CID 3825), diclofenac (CID 3033), indomethacin (CID 3715) and nicotine (CID 89594). https://pubchem.ncbi.nlm.nih.gov/
Flux figures in this article are calculated from the Potts–Guy permeability regression and PubChem aqueous solubility data for the neutral species. That regression was fitted to nonelectrolytes and carries roughly 0.7–0.8 log units of residual scatter; the NSAIDs are weak acids, so these are order-of-magnitude estimates outside the model’s fitted domain. They are used to show that the calculation fails to explain the marketed set, not to predict any product’s performance.
George Forrest is CEO and co-founder of Ibumix and a Chartered Chemical Engineer. He previously ran Stepan Company’s UK manufacturing site.
This article is an analysis of published literature and third-party marketed products. It is not medical advice, and it does not describe, promote or make any claim for any Ibumix product. Ibumix has no authorised medicinal product.
Ibumix is a UK pre-clinical pharmaceutical company developing a series of new liquid drug delivery platforms built on glyceryl caprylate/caprate (GCC) lipid chemistry. www.ibumix.com