A Comprehensive Evidence Review for Clinicians
Prepared: 2026-10-30 Style: Vancouver numbered citations, numbered by first appearance; every reference was generated mechanically from NCBI E-utilities metadata and every PMID resolved via esummary on the date of preparation Scope note: This review was built from primary literature retrieved through PubMed E-utilities, ClinicalTrials.gov v2, the Federal Register, and FDA/EMA regulatory pages on the date of preparation. Every cited record was retrieved and its title checked against the claim it supports.
Abstract
Transdermal drug delivery is the most familiar “advanced” delivery route in clinical medicine and the most tightly constrained. The stratum corneum permits useful passive flux only for small, moderately lipophilic, potent molecules — the reason the licensed patch pharmacopoeia is short and dominated by fentanyl, clonidine, scopolamine, nicotine, rivastigmine, rotigotine, lidocaine, nitroglycerin and a handful of hormones [1][2]. Metal–organic frameworks (MOFs) — crystalline coordination polymers of metal nodes and organic linkers — have been promoted as the platform that could lift that ceiling, on the strength of very high surface area, tunable porosity, modifiable surface chemistry and high loading capacity for hydrophobic drugs and macromolecules [3].
This review examines whether that promotion is warranted yet. Four findings organise the answer.
First, the mechanistic premise is sound but indirect. MOFs do not cross intact stratum corneum in useful quantities on their own; the credible routes to clinical relevance are hybrid: MOF–microneedle arrays that physically bypass the barrier, and MOF carriers that exploit hair follicles and intercellular lipids for local deposition [4][5]. Where MOF particles have been applied to skin, the dominant observed effect is increased cutaneous deposition, sometimes with reduced transdermal permeation — a distinction with direct clinical consequences that the literature frequently blurs [5][6].
Second, the evidence base is uniformly preclinical. Across 201 PubMed records at the intersection of MOFs and skin that we screened, and 48 of those abstracts read in full, 22 reported animal work and none reported a clinical study of a MOF-based transdermal therapeutic. ClinicalTrials.gov returns no registered trial of any MOF transdermal product: the five hits for “metal-organic framework” are unrelated records (a toothpaste trial, an H. pylori trial, an interstitial lung disease cohort, an olaparib trial and a traumatic brain injury study), and searches for ZIF-8, UiO-66, HKUST-1 and “metal-organic frameworks drug delivery” return zero relevant studies. The only MOF–microneedle device we identified with any human data is a diagnostic interstitial-fluid sensor, not a therapy [7].
Third, the safety questions are the right questions and remain unanswered. MOFs are designed to degrade; degradation releases metal ions and organic linkers whose local and systemic fate is measured in cell culture and rodents, rarely in human skin, and almost never after repeated application. Copper- and cobalt-based frameworks illustrate both the therapeutic appeal and the toxicological cost [8][9]. Regulatory frameworks exist for nanomaterials in drug products and for combination products, but they impose exactly the characterisation burden — particle fate, leaching, lot-to-lot reproducibility — that this literature does not yet supply [10][11].
Fourth, the adjacent technology that has actually reached humans is the dissolving microneedle patch, and it sets the realistic benchmark: randomised trials of influenza, measles–rubella and Japanese encephalitis vaccines delivered by microneedle patch have shown immunogenicity roughly comparable to subcutaneous injection with mild, transient local reactions [12][13][14]. No MOF has been tested in that setting.
The clinical bottom line is unambiguous: nothing here is prescribable, and nothing here is close. What would constitute real progress is equally unambiguous — human skin permeation and deposition data with mass balance, a repeat-dose local tolerability study, metal-ion pharmacokinetics, and a GMP-compatible manufacturing route. Until those exist, MOF transdermal delivery is a materials-science programme with a plausible hypothesis, not a therapeutic advance.
1. Introduction: Why the Skin Is Both the Opportunity and the Problem
1.1 The clinical appeal
Transdermal administration occupies a peculiar position in therapeutics: it is simultaneously mature and marginal. It is mature because every clinician has prescribed a patch and understands its virtues without needing a lecture — no first-pass hepatic metabolism, no dependence on gastric emptying or absorption, steady plasma concentrations rather than peaks and troughs, and a dosage form that a patient can remove if toxicity appears. It is marginal because the number of molecules that can actually use the route is small, and the reason is anatomical rather than pharmacological.
The seed review for this analysis frames the problem in exactly these terms: transdermal delivery avoids gastrointestinal degradation and hepatic first-pass metabolism, but is limited by the barrier function of the stratum corneum and by insufficient drug accumulation in the target tissue [3]. That framing is correct and is not in dispute anywhere in the literature. The disagreement — such as it is — concerns whether a new class of materials can change the second constraint without worsening the first.
1.2 The barrier, stated quantitatively
The stratum corneum is 10–20 µm of corneocytes embedded in a highly ordered lipid matrix dominated by ceramides, cholesterol and free fatty acids, arranged in stacked lamellae [15]. That architecture produces a diffusion barrier that is extraordinarily effective for hydrophilic solutes and for anything large. The practical consequence, stated plainly in a 2025 Nature paper describing a new skin-permeable polymer, is that non-invasive skin permeation “is widely used for convenient transdermal delivery of small-molecule therapeutics (less than 500 Da) with appropriate hydrophobicities” and “has long been deemed infeasible for large molecules — particularly polymers, proteins and peptides” [16].
That single sentence, from a paper whose entire point is to challenge it, is the best available summary of the constraint. The 500-dalton figure is not a physical law; it is an empirical regularity that has held well enough to define a pharmaceutical category. Permeability falls steeply with molecular size and is further modulated by lipophilicity, charge, melting point and the vehicle [17][18]. Databases of human maximum skin flux now contain enough measurements to build quantitative structure–permeation relationships, and those relationships consistently reproduce the same ordering: small and moderately lipophilic solutes cross; large or highly polar ones do not [18][19].
The clinical consequence is a short list. A 2024 review of transdermal systems names nicotine, fentanyl, nitroglycerin and clonidine as the canonical actives [1]; a dermatological review of patch reactions lists buprenorphine, diclofenac, lidocaine, oxybutynin, rotigotine, methylphenidate, rivastigmine, teriparatide and nicardipine among drugs newly available in patch form [2]. Every one of these is either small and lipophilic, or potent at low daily dose, or both. The exceptions prove the rule: teriparatide, a 34-amino-acid peptide, required a dedicated device programme to achieve fractionally useful delivery, and its appearance in a list of patch-form drugs is itself a statement about how hard the exception was to earn.
1.3 Three generations of enhancement, and where MOFs sit
The standard taxonomy divides transdermal enhancement into three generations [20]. First-generation systems rely on passive diffusion of small, lipophilic, low-dose drugs and account for the commercial patch market. Second-generation systems add chemical permeation enhancers, non-cavitational ultrasound or iontophoresis; several reached clinical products, and iontophoresis in particular allows real-time control of delivery rate. Third-generation systems deliberately breach or bypass the stratum corneum — microneedles, thermal ablation, microdermabrasion, electroporation, cavitational ultrasound — and it is here that macromolecules and vaccines become feasible [20].
MOFs are usually presented as a fourth thing: not an enhancement method but a carrier chemistry that can be bolted onto any of the three generations. That is both the appeal and the source of conceptual confusion in the literature. A MOF is not a permeation enhancer. It does not fluidise stratum corneum lipids or open a pore. Applied as a suspension or gel to intact skin, a 100–500 nm coordination polymer faces the same barrier as any other nanoparticle, and the evidence on nanoparticle penetration is discouraging: a review of 29 quantitative dermal penetration studies concluded that “skin is an efficient barrier for NPs, indistinctly of their properties,” with skin integrity and formulation more important determinants than particle characteristics, and — critically — that most studies used elemental analysis and therefore could not distinguish a dissolved fraction from intact particle penetration [6].
Where MOFs plausibly change the arithmetic is in three narrower roles:
- As a payload depot inside a microneedle. The microneedle breaches the barrier; the MOF controls loading and release. This is the dominant design in the recent literature and is explicitly described as the reason the hybrid exists — MOFs alone cannot penetrate skin, and microneedles alone have limited loading capacity [4].
- As a follicular delivery vehicle. Hair follicles provide a shunt past the stratum corneum, and at least one careful study shows ZIF-8 particles co-delivering drug and carrier into skin through follicular and intercellular routes [5].
- As a stimuli-responsive release element. Acid-labile frameworks such as ZIF-8 degrade in the mildly acidic environment of infected wounds, inflamed skin or tumour interstitium, providing a chemically triggered release that is genuinely difficult to achieve with a polymer depot [21][22].
None of these is a barrier-lifting mechanism in the sense that iontophoresis or microneedles are. Recognising that prevents most of the overclaiming that follows.
1.4 What this review does and does not do
This review asks four questions. What is the clinical baseline that any MOF platform must beat? What is the actual quality of the preclinical evidence, quantified rather than characterised? What are the credible failure modes — metal leaching, long-term biosafety, manufacturability, regulatory classification? And what would a clinician need to see before believing any of it?
It does not attempt a systematic review with PRISMA methodology, because the underlying literature does not support meta-analysis: the studies are heterogeneous in framework, cargo, model, dose and outcome, and almost none report the parameters needed to pool. It does not cite the seed review for anything quantitative, because the seed review reports no quantitative results in its abstract — it is a narrative review in a chemistry journal, and its contribution is framing rather than data [3]. And it does not describe the field as closer to translation than it is. The single most important sentence in this review is that no MOF-based transdermal product has been tested in a human trial of therapeutic efficacy, a claim supported below by direct registry searching rather than by inference.
2. The Clinical Baseline: What an Approved Transdermal System Achieves, and What It Costs
2.1 The pharmacopoeia and its pharmacology
Transdermal patches work because they maintain a concentration gradient across skin for hours to days, using either a drug-in-adhesive matrix or a rate-controlling membrane. The historical development is well documented and unremarkable: successive refinement of adhesive chemistry, backing films and release liners, with the active ingredient almost always selected to fit the route rather than the route engineered to fit the ingredient [23].
The clinical characteristics that matter to a prescriber are these:
- Steady-state kinetics without first-pass metabolism. This is the route’s genuine pharmacological advantage and the reason it survives in drugs with narrow therapeutic indices.
- Dose ceiling imposed by flux. A patch can only deliver what skin will pass. Typical maximum human skin fluxes for approved actives sit in the range that supports microgram-to-low-milligram daily doses; the human flux database assembled from in vitro permeation testing makes this quantitatively explicit across hundreds of solutes [18].
- Dose-dumping risk. A defective matrix or a breached backing can release the entire reservoir, which is why patch formulations are subject to specific quality and failure-mode scrutiny [1].
- Cutaneous intolerance as a class effect. Irritant and allergic contact dermatitis are common enough that a dedicated dermatological literature exists on the subject, covering buprenorphine, diclofenac, lidocaine, oxybutynin, rotigotine, methylphenidate, rivastigmine, teriparatide and nicardipine [2]. Any new patch platform inherits this tolerability burden.
2.2 What “beating the baseline” would mean numerically
This is where the MOF literature is weakest, and it is worth being precise about why. A new transdermal platform can improve on the baseline in exactly four ways:
| Improvement axis | What must be measured | What the baseline already achieves |
|---|---|---|
| Extend the molecular size range | Flux (µg/cm²/h) for a molecule >500 Da, with mass balance | Largely closed above ~500 Da for passive delivery, per a source that describes useful passive delivery as confined to <500 Da molecules [16] |
| Increase flux for an existing drug | Comparative flux vs licensed formulation, same skin model | Established for each approved active [18] |
| Improve tolerability | Repeat-application irritation/sensitisation rates in humans | Irritation is common but usually mild [2] |
| Enable a new cargo class | Proof of systemic exposure, not just deposition | Not achieved passively for proteins [16] |
Almost every MOF paper in this review claims the first or fourth axis. Very few report data in the form the table requires. The most common substitute is skin deposition — the amount of drug recovered from homogenised skin after an experiment — which is a legitimate measurement but answers a different question. A platform that increases deposition tenfold while leaving systemic absorption unchanged is a topical product, not a transdermal one, and the two are regulated, priced and prescribed differently.
2.3 The deposition-versus-absorption distinction, made concrete
The clearest demonstration in the MOF literature comes from a ZIF-8/polydopamine study of 5-fluorouracil delivery [5]. The authors used Franz cells with ex vivo skin, confocal microscopy and histology to trace both drug and carrier. Two findings deserve emphasis. First, both ZIF-8 and ZIF-8@polydopamine increased 5-FU deposition in skin. Second — and this is the part usually omitted when the paper is cited — the polydopamine coating reduced drug permeation across the skin, “especially in the case of impaired skin,” compared with a plain drug solution [5]. For a dermatological indication such as field treatment of actinic keratosis or a superficial basal cell carcinoma, reduced systemic exposure with increased local deposition is exactly the right outcome. For a systemic indication it is a failure.
The same paper also shows that the carrier itself can cross into skin as an intact particle via the follicular route, which raises a safety question rather than answering one: an intact coordination polymer in a sebaceous follicle is a foreign body in a location with its own inflammatory pathology.
This distinction is not a pedantic one. It determines whether a product would be developed as a topical dermatological, a transdermal systemic, or a local device — three regulatory pathways with different evidence requirements [11]. Any future paper that reports “enhanced transdermal delivery” without separating deposition from permeation should be read as reporting a topical effect until proven otherwise.
2.4 Franz cells, and why the model matters clinically
Nearly all MOF skin-permeation data are generated in Franz diffusion cells. The methodology is well standardised — OECD Test Guideline 428 and Guidance Document 28 define the conduct of in vitro skin absorption studies, and a 2020 consensus commentary from hands-on practitioners reviews their limitations in detail [24]. Three limitations are directly relevant to interpreting this literature.
Membrane choice changes the answer. Human cadaver epidermis, porcine skin, bovine skin, reconstructed human epidermis and rodent skin are not interchangeable. A direct comparison of carprofen permeation across human, porcine and bovine skin found similar diffusion for human and porcine but higher for bovine [25]. Reconstructed epidermis has been proposed as a replacement for excised skin but requires careful protocol transfer to be comparable [26]. Rodent skin — the default in most MOF papers, usually mouse or rat — is substantially more permeable than human skin, largely because of follicular density and lipid composition. A flux value obtained on mouse skin does not transfer.
Barrier integrity testing is itself uncertain. A retrospective analysis of 9,978 electrical-resistance measurements over 15 years found that the widely used 10 kΩ threshold for a 2.54 cm² cell identified half of all samples as potentially damaged, many of them false positives — meaning that studies applying that criterion may have been discarding valid skin, while studies applying no criterion may have included damaged skin [27]. Either way, cross-study comparison of absolute flux values is fragile.
Sink conditions and receptor phase determine what “permeation” means. Aqueous receptor phases under-represent lipophilic drugs; albumin-containing phases change free fraction; and short experiments truncate slow pathways. The OECD documents address these points, but adherence varies [24].
There is a further methodological layer specific to this field: most MOF papers report permeation of a surrogate. Confocal microscopy with rhodamine 6G as a fluorescent 5-FU mimic, for instance, tells you where a fluorophore goes, not where the drug goes [5]. Fluorescent tags can dissociate from the carrier, and free dye has different partitioning from a loaded framework.
None of this makes the existing data worthless. It makes it non-transferable — which is why in vitro–in vivo correlation for transdermal systems is a recognised regulatory discipline in its own right, with its own modelling approaches and its own translational failures [28].
2.5 Adverse-event burden of the existing route
A new platform is not competing against an abstract barrier; it is competing against a route with a known and tolerable safety profile. Cutaneous reactions to transdermal systems are common, usually irritant, sometimes allergic, and rarely severe [2]. For a clinician, this sets the tolerability bar: a MOF patch that produces more than transient erythema in a meaningful fraction of recipients would be worse than what is already available, before any question of efficacy arises. The microneedle literature, discussed in Section 10, provides the relevant comparator rates.
2.6 Why the baseline matters for judging MOFs
The purpose of this section is not to celebrate existing patches. It is to establish that the target is well defined and measurable. A field that reports “significantly promoted wound closure” in diabetic mice against an untreated control is not measuring itself against the baseline that matters. It is measuring itself against nothing. The baseline that matters is a licensed product, in human skin, with a flux or a clinical endpoint, and with a tolerability profile. Sections 4 through 7 should be read with that scorecard in hand.
3. What MOFs Are, and Which Properties Actually Motivate Transdermal Use
3.1 Definition and structural logic
A metal–organic framework is a crystalline porous solid in which metal ions or clusters (“nodes”) are connected by multitopic organic ligands (“linkers”) into a periodic three-dimensional lattice. The defining features are permanent porosity, structural periodicity and chemical tunability: because both node and linker can be varied almost indefinitely, the framework’s pore size, surface chemistry, hydrophobicity and degradation behaviour can in principle be engineered.
The number of known frameworks is enormous. One 2023 review notes that “nearly ninety thousand MOFs have been investigated” [29], and a 2025 Nature Communications paper reports that “over 5000 new metal–organic frameworks were reported in the past year alone” [30]. That scale is the field’s principal strength — a vast design space — and its principal weakness, because synthesis and characterisation capacity cannot keep pace with the combinatorial possibilities, and the biomedical subset that has actually been tested in biological systems is a tiny and non-random fraction.
3.2 The four properties that matter for skin
The seed review lists high surface area, tunable pore structure, versatile surface chemistry and high drug-loading capacity, particularly for hydrophobic compounds and macromolecules, plus programmable multidrug delivery and integration of diagnosis with therapy [3]. Each of these deserves a clinician’s translation.
High surface area and porosity. BET surface areas for canonical frameworks run from hundreds to several thousand m²/g. In practical terms this means a small mass of carrier can adsorb a large mass of drug. A Zr-based framework, SU-102, loaded ciprofloxacin to 33.3% w/w with 66.8% efficiency by ion exchange [31]; a mesoporous Zr framework (NU-1000) immobilised insulin at approximately 40 wt% within 30 minutes [32]; a cyclodextrin framework carried more than 26% w/w quercetin [33]. These are genuine and unusually high loading figures by the standards of polymeric nanocarriers, and they matter because a microneedle has a finite volume: the more concentrated the payload, the fewer needles are needed and the smaller the patch.
The caveat is that high loading capacity measured by thermogravimetric analysis or UV spectroscopy does not establish that the drug is released in an active form at the target site. Loading is necessary but not sufficient.
Tunable pore structure and surface chemistry. This is what allows a framework to be selected for a cargo rather than the reverse. Hydrophobic drugs that are poorly soluble in water can be accommodated in hydrophobic pores; polar cargoes can be attached to functionalised linkers; and pore apertures can be sized to exclude large molecules. The practical expression of this in the transdermal literature is the cyclodextrin MOF, where the chiral, cavity-bearing linker both hosts the drug and improves aqueous solubility — one study reported a 267.1-fold increase in curcumin solubility in water with 13.9% drug loading [34].
Programmable multidrug delivery. Sequential release of two agents with different kinetics is difficult with a single polymer and straightforward with a core–shell microneedle containing two different matrices. A representative example encapsulates CuO₂ in ZIF-8 within a pH-sensitive hyaluronic acid core for rapid release in an acidic wound, and VEGF in a gelatin-methacrylate shell for sustained release [35]. Whether this is better than two separate dressings is a clinical trial question that has not been asked.
Degradability as a feature. The property most often marketed as biocompatibility is really controlled instability. ZIF-8 hydrolyses in acid, releasing zinc and 2-methylimidazole; this is described in the review literature as “intrinsic pH-responsive biodegradability” [21] and is exploited deliberately in glucose-responsive insulin delivery, where glucose oxidase generates gluconic acid locally, lowering pH and degrading the framework to release insulin [22]. The same property is the central safety liability, discussed in Section 8.
3.3 What MOFs are not
Three clarifications prevent most errors of interpretation.
They are not nanoparticulate drugs with a defined pharmacological profile. A MOF is a material, and its behaviour depends on particle size, morphology, surface coating, dispersion state and the biological medium. Two papers using “ZIF-8” may be studying physically quite different objects. This is why the toxicity literature is so difficult to synthesise: dose is reported variously as µg/mL, mg/L, µg/cm² or mg/kg, and particle characterisation is often incomplete [36].
They are not automatically biocompatible. The seed review lists “favourable biocompatibility” among claimed advantages [3], and many individual papers report good cytocompatibility in one or two cell lines. But a dedicated tutorial review on MOF nanoparticle toxicity argues that systematic assessment is the exception rather than the rule, and that the field needs to move toward “safe-and-sustainable-by-design” synthesis [36]. A second review notes that although tens of thousands of frameworks exist, evaluating their properties “in relevant biological systems remain[s] critical for their clinical translation” [29]. Cytocompatibility in HaCaT cells at 24 hours is a starting point, not a conclusion.
They are not a delivery route. The single most important structural fact in this review is that a MOF applied to intact skin does not meaningfully cross it. The 2025 review of MOF microneedles states the logic without hedging: “MOFs alone cannot penetrate the skin or mucosal barrier to deliver drugs effectively, and MNs have limited direct loading capacity” [4]. Everything that follows depends on which partner technology supplies the barrier breach.
3.4 The synthesis-to-application gap
There is a further structural problem that affects how the literature should be read. MOF synthesis is usually reported as a bespoke batch process optimised for a single paper: solvothermal reaction in dimethylformamide at 120 °C for 24 hours, or a room-temperature aqueous precipitation for ZIF-8. Medical application requires the opposite — a reproducible, scalable, impurity-controlled process in a solvent acceptable to a pharmaceutical regulator.
The scale-up literature exists but is thin and largely non-pharmaceutical. A representative example is a microwave-assisted synthesis of UiO-66 optimised across a wide parameter space and scaled while maintaining efficiency, with the authors reporting no significant effect on morphology, crystal size or defects compared with conventional heating [37]. That is a genuine contribution to manufacturability. It is not a GMP process, and it does not address residual solvent, linker purity, metal source traceability, or the endotoxin burden that any parenteral or implanted product would require.
A 2025 multimodal machine-learning paper makes the structural point well: frameworks are typically synthesised for one application and may be suited to entirely different ones, but “linking these new materials to their best applications remains a significant challenge” [30]. In the transdermal case, the field has largely skipped that linking step and moved directly from synthesis to in vivo wound-healing claims.
3.5 Why the transdermal framing is attractive to materials scientists — and what clinicians should take from it
The scientific appeal is real. Skin is accessible, the experiment is cheap, the readout (wound photography, histology, fluorescence) is visually compelling, and a positive result is easy to publish. A materials chemist with a new framework and a diabetic mouse model can produce a complete story in months. That incentive structure explains the shape of the literature better than any scientific argument does: 153 of the 201 MOF–skin records we screened were published in 2024 or later, and the modal paper is a new framework plus a wound-healing model plus a mechanism story.
For a clinician, the correct inference is not that the work is fraudulent or worthless. It is that the field’s selection pressure has favoured novelty of material over rigour of translation, and that the resulting corpus cannot yet answer the questions a prescriber would ask. Sections 4 through 7 describe what has actually been built. Section 12 quantifies how far it is from a product. ## 4. MOF Classes in Transdermal Work: Composition, Stability and the Metal-Ion Question
4.1 Why class matters more than the label “MOF”
Papers in this field routinely say “a MOF was synthesised” as though frameworks were interchangeable. They are not. The metal determines the biological activity and the toxicity; the linker determines the degradation chemistry and the organic burden; the topology determines whether the drug fits and how fast it leaves. A prescriber reading “MOF patch” should mentally substitute “a specific metal salt plus a specific organic acid, held in a lattice that will fall apart somewhere in the skin.”
The table below summarises the classes that actually appear in transdermal and cutaneous work, with their composition, the degradation products that matter and representative transdermal uses. Composition is standard coordination chemistry; the degradation and application claims are cited.
| Class | Metal / linker | Degradation products | Representative cutaneous/transdermal work |
|---|---|---|---|
| ZIF-8 | Zn²⁺ / 2-methylimidazolate | Zn²⁺, 2-methylimidazole | 5-FU deposition [5]; minocycline in silk microneedles for acne [38]; glucose-responsive insulin [22]; wound dressings [39] |
| ZIF-67 | Co²⁺ / 2-methylimidazolate | Co²⁺, 2-methylimidazole | Neurotoxicity signal in rats at non-general-toxic doses [9] |
| HKUST-1 / Cu-BTC | Cu²⁺ / benzene-1,3,5-tricarboxylate | Cu²⁺, trimesic acid | Nitric oxide release for diabetic wounds [40][41]; folic-acid-stabilised wound healing [42]; renal cytotoxicity [8] |
| UiO-66 / UiO-67 | Zr⁴⁺ / terephthalate or biphenyl-dicarboxylate | Zr⁴⁺, terephthalic acid | Sonodynamic antibacterial wound therapy [43]; 5-FU transdermal hydrogel [44]; amoxicillin scaffold [45] |
| MIL-100(Fe), MIL-89 | Fe³⁺ / trimesate or muconate | Fe³⁺, carboxylic acids | Dacarbazine against melanoma cells [46]; cosmetic caffeine/ibuprofen patches [47]; low cytotoxicity in hepatocytes [48] |
| Zr-MOF-808, SU-102 | Zr⁴⁺ / trimesate or related | Zr⁴⁺, carboxylates | Proinflammatory mediator adsorption in scars [49][50]; ciprofloxacin release [31] |
| Cyclodextrin MOFs (CD-MOF) | K⁺ / γ-cyclodextrin | Cyclodextrin, K⁺ | Curcumin solubilisation and topical delivery [34]; quercetin for hypertrophic scar [33]; intrathecal methylprednisolone [51] |
| Bio-MOFs / Zn-adeninate, Zn-curcumin | Zn²⁺ / nucleobase or curcumin | Zn²⁺, linker | Curcumin permeation across biomimetic membranes [52] |
4.2 ZIF-8: the workhorse, and why
ZIF-8 dominates the transdermal and wound literature for three reasons: it assembles in water at room temperature from cheap reagents, it has a high loading capacity, and it is acid-labile in a convenient pH range. The review literature describes its “intrinsic pH-responsive biodegradability, adjustable porosity and surface area, inherent antibacterial activity, and excellent biocompatibility” as the basis of its appeal [21]. Zinc is also an established wound-healing ion, which allows authors to present framework degradation as therapeutic rather than as a liability.
That framing deserves scrutiny. Zinc is not biologically inert — its therapeutic and cytotoxic concentration ranges overlap — and 2-methylimidazole is not a physiological molecule. The typical paper reports cytocompatibility at the concentrations used, which is reassuring but narrow: it establishes that a particular dose of a particular preparation did not kill a particular cell line within 24–48 hours. It does not establish what happens after daily application for eight weeks to intact or broken skin.
The most instructive ZIF-8 transdermal study is the 5-FU work already discussed, because it measured the right things: Franz-cell permeation, confocal localisation and histology [5]. Its finding that the polydopamine-coated framework increased deposition while reducing permeation across impaired skin is a genuinely useful result — it defines a topical product candidate, not a systemic one. The acne study is similarly well specified, reporting needle failure force of approximately 0.4 N, dissolution under one minute, drug loading of approximately 40.11% and encapsulation efficiency of approximately 70.2%, with faster release at pH 5.5 than at pH 7.4 [38]. Those are the numbers a formulation scientist needs. They are not the numbers a clinician needs.
4.3 Copper frameworks: the most interesting and the most concerning
Copper-based frameworks, principally HKUST-1 (also called Cu-BTC), occupy a distinctive position because copper is both a therapeutic and a toxicant. Copper ions promote angiogenesis and are antibacterial; copper excess is hepatotoxic and nephrotoxic. The literature has effectively built a case that controlled copper release is beneficial, and a parallel case that uncontrolled copper release is harmful — and both are correct.
On the therapeutic side, a copper MOF embedded in a thermoresponsive citrate hydrogel released copper slowly, reduced cytotoxicity and apoptosis compared with free copper, and increased wound closure in diabetic mice [53]. Folic-acid modification of HKUST-1 slowed copper release further, reduced cytotoxicity, and improved closure in the same model [42]. A HKUST-1 core in coaxial nanofibres released nitric oxide at an average rate of 1.74 nmol/L/h for more than 14 days [40] — one of the few genuinely quantitative release measurements in this corpus.
On the toxicological side, HKUST-1 produced concentration-dependent cytotoxicity in human embryonic kidney cells, with mitochondrial membrane depolarisation, apoptotic bodies, autophagic vesicles, oxidative stress, NF-κB and NLRP3 activation, and depletion of zonula occludens proteins — that is, loss of the barrier integrity of a renal epithelial monolayer [8]. This is an in vitro study, and the concentrations used may exceed anything achievable in vivo from a skin patch. But it identifies a specific, plausible mechanism of harm for a specifically plausible off-target organ, and no transdermal study has excluded it.
The unresolved question is mass balance. A 1 cm² patch containing 3% w/w Cu@ZIF-8 delivers an amount of copper that can be calculated; what fraction enters the systemic circulation, what fraction stays in skin, and what the steady-state serum copper is over 28 days have not been reported for any transdermal MOF. For a wound dressing this may be acceptable, because the wound is open and the duration is bounded. For a cosmetic or a chronic-use patch it is not.
4.4 Zirconium frameworks: stability as the selling point
Zr-based frameworks (UiO-66, UiO-67, MOF-808, SU-102) are among the most hydrolytically and thermally stable MOFs known, which makes them attractive when the application requires the framework to persist. Two consequences follow.
First, persistence can be the therapeutic mechanism. MOF-808 crystals were shown to remove nitric oxide, cytokines and reactive oxygen species by adsorption rather than by pharmacology — a “multiscale adsorption mechanism” — and a silicone patch loaded with 0.2% MOF-808 reduced scar elevation in a rabbit ear model, with TGF-β1, collagen-1 and α-SMA levels lower by 8.8%, 12% and 21.3% respectively in the treated group [49][50]. That is an unusual and interesting mechanism: the material acts as a molecular sponge. It is also the mechanism most likely to raise regulatory questions, because an adsorbent that persists in tissue is closer to an implanted device than to a drug.
Second, persistence is a safety question. Zirconium is not an essential element and has no dedicated physiological clearance pathway, so a non-degrading framework has no obvious route out of tissue. A framework designed not to degrade must be shown to be eliminated or to be harmless where it stays. No transdermal study does this.
UiO-66 also appears in sonodynamic therapy, where ultrasound-activated framework generates reactive oxygen species and, in a diabetic mouse model of S. aureus-infected wounds, reduced bacterial burden and accelerated closure [43]. The mechanistic claim — that a pristine, unmodified framework can act as a sonosensitiser — is interesting; the translational distance is the same as everywhere else in this field.
4.5 Iron frameworks: the biocompatibility argument
MIL-100(Fe) and related iron carboxylates are attractive because iron is endogenous and the linker is a simple tricarboxylic acid. An in vitro toxicity study in human hepatocytes and hepatoma cells reported a “safe dose” of 80 µg/mL with cell viability above 85.97% and 91.20% respectively [48] — a modest but honest result. MIL-100(Fe) has been used for dacarbazine delivery against melanoma cells [46] and, notably, in one of the earliest cutaneous MOF formulations: a press-moulded polymer–MIL-100(Fe) composite patch loaded with caffeine and ibuprofen, which achieved very high caffeine loading and progressive release, with permeation reaching adipose tissue in an ex vivo model [47].
That 2016 paper is worth flagging for a historical reason. It is one of the few in this corpus that was designed as a cutaneous product — a patch for cosmetic lipolysis and analgesia — rather than as a wound-healing demonstration. Nine years later, the field has produced no clinical candidate from that line. The absence is informative.
4.6 Cyclodextrin frameworks: the most pharmaceutically plausible
Cyclodextrin MOFs deserve separate treatment because they are the class closest to pharmaceutical convention. The linker is γ-cyclodextrin, a GRAS excipient already used in marketed products; the metal is potassium; and the framework is assembled in water. A 2025 review of CD-MOF design principles identifies the key limitations candidly: “aqueous instability, scalability, and precise functionalization hinder their full potential” [54].
The transdermal results are among the more credible in the field. A γ-CD-MOF achieved 13.9% curcumin loading and a 267.1-fold increase in aqueous solubility, and a composite with cyclodextrin nanosponges improved in vitro release and transdermal permeation [34]. A diphenyl-carbonate-crosslinked CD-MOF loaded more than 26% w/w quercetin and, coated with a hypertrophic-scar-fibroblast membrane and delivered by dissolvable microneedles, reduced collagen I and III expression and modulated Wnt/β-catenin and JAK2/STAT3 signalling in a scar model [33]. An epidural microneedle array using CD-MOF for methylprednisolone delivery after spinal cord injury improved motor scores and reduced M1 microglia in animals [51] — a reminder that “transdermal” in this literature sometimes means “through the dura,” which is a different product category entirely.
4.7 The metal-ion question, stated plainly
Every MOF degrades into its constituents eventually. For transdermal use the relevant questions are: how fast, into what, at what local concentration, and with what systemic exposure. The literature answers the first question variably, the second question by assertion, the third almost never, and the fourth essentially never.
What is known about metal ion permeation through skin is not reassuring in the abstract. A 2025 Franz-cell study of gallium arsenide particles found that arsenic concentrations in the receptor compartment after 24 hours were three times higher in damaged skin (1,558 ± 546 ng/cm²) than in decontaminated skin (458 ± 144 ng/cm²), and gallium concentrations six times higher (244 ± 5.29 vs 37.1 ± 11.9 ng/cm²) [55]. Intact skin was an effective barrier; damaged skin was not. Diabetic ulcers, eczema, tape-stripped skin, microneedle-created micropores and inflamed acne lesions are all damaged skin. The relevance to a zinc- or copper-releasing framework applied to exactly those lesions is direct and has not been studied.
5. Mechanisms and Enhancement Strategies
5.1 Passive delivery: what the data actually show
Passive application of MOF particles to skin produces measurable deposition, occasionally measurable permeation, and rarely systemic absorption. The mechanisms invoked are intercellular diffusion through the lipid lamellae, the follicular shunt, and transappendageal transport. The ZIF-8/polydopamine study used confocal microscopy with a fluorescent probe and concluded that both intercellular and follicular routes operated, with the follicular route “particularly important” for the coated particles, which were co-delivered into skin as intact particles [5].
That result cuts both ways. Follicular delivery is genuinely useful for acne, alopecia and follicular disorders. It is also a route into a compartment with a rich immune population and its own disease spectrum, and there is no chronic safety data for particulate accumulation in sebaceous follicles.
One further passive strategy deserves mention because it is often claimed but rarely demonstrated: molecular permeation enhancement by the framework itself. In the bio-MOF curcumin study, the carrier “successfully penetrated the biomimetic barrier and released Cur in the receptor compartment” [52] — but the barrier was a biomimetic membrane, not skin, and the readout was fluorescence. A biomimetic membrane is a useful screening tool; it is not evidence of transdermal delivery.
5.2 Microneedle–MOF hybrids: the dominant architecture
The logic of the hybrid is stated explicitly in the 2025 review of MOF microneedles: MOFs cannot cross skin, microneedles cannot carry much, and together they solve each other’s problem — the framework raises loading efficiency and modulates release, while the microneedle provides the breach and improves mechanical properties [4]. This is the most defensible design in the field and accounts for the majority of the recent high-profile papers.
Three subtypes appear:
Dissolving microneedles. The needle matrix itself dissolves in interstitial fluid, releasing the loaded framework. Examples include silk fibroin needles carrying minocycline-loaded ZIF-8 for acne, with dissolution under one minute [38]; polyvinyl alcohol/alginate/maltose needles carrying Cu@ZIF-8 for full-thickness wounds, where adding maltose increased compressive force from 0.03 to 0.143 N per needle and 3% Cu@ZIF-8 raised it further to 0.156 N [39]; and γ-polyglutamic acid needles carrying metformin-loaded Ga-Car-MOF in a double-layer patch [56].
Hydrogel-forming microneedles. The needle swells rather than dissolves, creating a hydrogel conduit through which the framework diffuses. These are better suited to sustained release and to repeated use, and they are the subtype most often paired with photothermal or stimuli-responsive elements [57].
Core–shell and separable designs. Two matrices with different degradation kinetics allow programmed sequence. The clearest example delivers CuO₂@ZIF-8 from a pH-sensitive hyaluronic acid core for rapid antibacterial action, then VEGF from a gelatin-methacrylate shell for angiogenesis — infection control followed by repair, in that order, from one application [35].
Mechanical performance is reported more consistently than in the rest of the field, because a microneedle that cannot penetrate skin fails visibly. Failure forces in the 0.1–0.4 N per needle range recur across papers [38][39], which is in the range needed to penetrate stratum corneum. That is a genuine, if narrow, piece of translational progress: the physical device works.
What has not been shown is that the framework survives the manufacturing process intact, that the needles dissolve completely in human skin rather than rodent skin, or that the released particles behave as they did in vitro. A 2024 review makes the general point sharply: “no nanoparticle-loaded DMN products have been approved thus far” and “the lack of understanding regarding their in vivo fate represents a critical bottleneck” [58].
5.3 Iontophoresis and sonophoresis: an honest negative
This is one of the more striking findings of this review. Iontophoresis is an established second-generation enhancement technology with licensed products and real-time controllability [20]. Sonophoresis has a 30-year evidence base, including a 1995 Science paper showing that low-frequency ultrasound increased human skin permeability to proteins including insulin, interferon gamma and erythropoietin by several orders of magnitude [59]. Combining either with a MOF would be an obvious research programme.
It has essentially not been done. A PubMed search combining “metal-organic framework” with iontophoresis, sonophoresis or electroporation returned three records, none of which is a MOF transdermal delivery study: one on irreversible electroporation in pancreatic cancer, one on MOF-based wearable sensors, and one on triboelectric disinfection. There is no published MOF–iontophoresis transdermal platform.
What exists instead is sonodynamic therapy, which uses ultrasound to activate a sonosensitiser rather than to permeabilise skin — a different mechanism with a different energy dose. UiO-66 activated by ultrasound generates reactive oxygen species and accelerates infected diabetic wound closure in mice [43], and a hollow porphyrinic MOF carrying oxygen-enriched perfluorooctyl bromide was integrated into dissolving microneedles for sonodynamic treatment of hypertrophic scar, with the microneedle providing the barrier breach and the ultrasound providing activation [60].
The distinction matters clinically. Sonophoresis is a delivery technology that could in principle be evaluated against existing transdermal systems. Sonodynamic therapy is a local ablative-adjuvant technology competing with intralesional injection and laser therapy. Conflating them in a review would overstate the maturity of the delivery science.
5.4 Photothermal and light-triggered release
Near-infrared photothermal control is the most developed external trigger in this literature, because it is easy to implement and visually demonstrable. HKUST-1 coated with graphene oxide and embedded in a porous PEGDA microneedle released nitric oxide under NIR irradiation and accelerated vascularisation, tissue regeneration and collagen deposition in type 1 diabetic rats [41]. A dual-layer patch combined mild photothermal therapy with autophagy regulation via an Hsp70 inhibitor for triple-negative breast cancer, which is a transdermal route to a non-skin disease [57]. Copper-MOF/polydopamine composites exploit d-d transitions to enhance photothermal conversion and generate hydroxyl radicals in the acidic tumour microenvironment [61].
The clinical question for any photothermal patch is dose and safety of the light, not just of the drug. NIR at the irradiances used in these papers requires controlled application; a home-use patch with a light source is a device with its own regulatory pathway and its own burn risk.
5.5 Chemical and biological triggers
pH. The dominant trigger, because infected wounds, tumours and inflamed skin are acidic and ZIF-8 is acid-labile. Release at pH 5.5 substantially exceeded release at pH 7.4 in the minocycline/ZIF-8 silk microneedle [38], and glucose oxidase-generated gluconic acid drives framework degradation in the insulin system [22].
ROS. Polydopamine-coordinated copper MOFs released dictamnine in response to reactive oxygen species, with superoxide radical scavenging of 63.85 ± 0.34% and encapsulation efficiency of 84.62 ± 2.14%, in a separable microneedle for atopic dermatitis [62].
Enzyme. Framework protection of enzymatic cargo is a related but distinct use: Ca-MOF shielded Clostridium histolyticum collagenase from degradation while releasing Ca²⁺ and gallic acid to modulate fibroblasts in a scar model, with needles dissolving within 25 minutes [63].
Magnetic. Iron oxide nanoparticles confined within a framework produced magnetothermally triggered release, addressing the aggregation problem that limits conventional magnetic carriers [64]. This is elegant chemistry with no cutaneous application yet.
A 2026 review of stimuli-responsive polymer–MOF composites provides the best available synthesis of this space and is candid that systematic comparison of mechanisms and performance across systems has not been done [65].
5.6 What the mechanism literature does not establish
Across all five categories, three things are missing.
First, dose–response. Almost no paper reports a dose-ranging study in vivo. The comparison is treated versus untreated, occasionally versus a free-drug control at a single dose. Without dose–response, “the framework improves delivery” cannot be separated from “more drug was applied.”
Second, mechanism discrimination. When a MOF microneedle improves wound healing, is the active agent the drug, the metal ion, the linker, the microneedle-induced microtrauma, or the framework’s adsorption of inflammatory mediators? Several designs could plausibly work with the framework omitted, and the control that would answer this — same device, empty framework — is inconsistently reported.
Third, human skin. Almost all permeation data are from rodent or porcine skin, and almost all efficacy data from rodents. Where human skin is used, it is usually cadaver epidermis in a Franz cell, which answers a permeation question but not a tolerability or efficacy one.
6. Cargo Classes: Small Molecules, Macromolecules, Nucleic Acids and Vaccines
6.1 Small molecules: the strongest and least interesting evidence
Small-molecule cargo dominates the literature because it is easy: the drug is stable, soluble in the right solvent, and quantifiable by HPLC or UV. Representative examples, with the reported loading and release data:
| Cargo | Framework / device | Reported metric | Source |
|---|---|---|---|
| 5-Fluorouracil | ZIF-8 ± polydopamine, topical | Increased skin deposition; reduced permeation with PDA coating | [5] |
| 5-Fluorouracil | IL@UiO-66-NH₂ hydrogel | Loading 671 mM (G1) and 397.8 mM (G2); 76.4% and 82.7% penetration over 48 h | [44] |
| Curcumin | γ-CD-MOF + nanosponge | 13.9% loading; 267.1-fold solubility increase | [34] |
| Curcumin | Zn bio-MOF | 75.98 ± 2.65% loading efficiency; permeation across biomimetic membrane | [52] |
| Quercetin | CD-MOF, HSF-membrane coated, microneedle | >26% w/w loading | [33] |
| Minocycline | ZIF-8 in silk microneedle | ≈40.11% loading; ≈70.2% EE; pH 5.5 accelerated release | [38] |
| Ciprofloxacin | Zr MOF SU-102 | 33.3% loading; 66.8% efficiency | [31] |
| Taxifolin | HKUST-1 hydrogel | 41.94 ± 2.60% loading | [66] |
| Dacarbazine | PEGylated MIL-100(Fe) | Cytotoxicity vs A375 melanoma cells | [46] |
| Caffeine, ibuprofen | MIL-100(Fe) polymer patch | “Very high” caffeine loading; progressive release | [47] |
Two patterns are worth naming. Loading capacities of 30–40% w/w are routine and genuinely impressive; they exceed what liposomes or polymeric micelles typically achieve and they justify the framework’s claim to be a high-capacity depot. Second, the permeability data are almost always from a non-human, non-intact or non-physiological model — biomimetic membranes, rat skin, or ex vivo tissue with the stratum corneum compromised.
The 5-FU hydrogel result is the most quantitatively complete transdermal dataset in the corpus: 76.4% and 82.7% of applied drug penetrating over 48 hours, in a formulation with high loading and ∼90% HaCaT viability [44]. Even here, the membrane is not specified as human skin in the abstract, and 48-hour penetration percentages are not the same as steady-state flux.
6.2 Macromolecules: the real prize, and the weakest data
Proteins and peptides are the cargo class where transdermal delivery would change practice most — insulin without needles, growth hormone without injections, antibodies without infusions. They are also the class where the barrier is most absolute: the 2025 Nature polymer paper states plainly that transdermal delivery of “polymers, proteins and peptides” has “long been deemed infeasible” [16].
MOF-based macromolecule delivery has been demonstrated, but not transdermally. Insulin was loaded into mesoporous NU-1000 at approximately 40 wt% in 30 minutes and protected from gastric acid and pepsin — an oral delivery study [32]. ZIF-8 encapsulation preserved insulin’s chemical stability and structural integrity against heat, agitation and organic solvent, and preserved bioactivity after storage at 40 °C in a diabetic mouse model [67]. ZIF-8 encapsulation likewise preserved and photothermally enhanced enzyme activity [68]. These are real and useful contributions to biologic stabilisation.
The transdermal macromolecule examples are narrower. Glucose-responsive insulin delivery used insulin and glucose oxidase co-loaded into cobalt-doped ZIF-8 within microneedles, with release driven by glucose oxidase-generated acid and free Co²⁺ chelated by EDTA-functionalised silica to limit systemic exposure [22]. That is the most clinically imaginable design in the corpus — a patch that responds to a physiological signal — and it has never been tested in a large animal, let alone a human. Nerve growth factor was loaded into Ce-UiO-66 for diabetic ulcer healing, with RNA sequencing implicating NF-κB suppression and recovery of skin neuroendocrine function [69]. Collagenase was protected by Ca-MOF for scar remodelling [63]. VEGF was released from a microneedle shell for angiogenesis [35].
Note what is absent: no pharmacokinetic data for any macromolecule delivered transdermally by a MOF. No serum concentration–time curve. No bioavailability estimate. In a field where the central claim is “we can deliver proteins through skin,” the absence of a single plasma insulin or plasma NGF profile is the most consequential gap in this review.
6.3 Nucleic acids: proof of concept in the wrong model
Nucleic acid delivery by MOF is at an earlier stage still. A 2026 review of nano/micro-scale RNA delivery for cancer lists MOFs among the vector classes alongside lipid nanoparticles, polymeric micelles, hydrogels, microneedles, exosomes and covalent organic frameworks, without identifying a transdermal MOF-RNA product [70].
The most concrete nucleic-acid work in the corpus is not therapeutic at all. A microneedle-templated UiO-66 coating captured and released DNA in situ, exploiting unsaturated Zr sites for selective nucleic acid accommodation and elution, and survived cyclic immersion [71]. That is a sampling technology — useful for biopsy-free molecular diagnostics, unrelated to gene therapy.
For contrast, the microneedle field without MOFs has already delivered mRNA. A dissolving microneedle carrying an mRNA-liposome vaccine preserved mRNA integrity during fabrication, promoted antigen-presenting cell migration to draining lymph nodes within 24 hours, and generated antigen-specific T-cell responses in mice [72]. The framework adds nothing to that problem and may add stability and regulatory risk.
6.4 Vaccines: the gap between the adjacent field and this one
Transdermal immunisation is the clearest example of what this technology class could become, and the clearest demonstration that MOFs are not yet part of it.
The microneedle vaccine literature has produced randomised human trials with immunogenicity endpoints. A phase 1 trial of dissolving microneedle influenza vaccine in 100 participants found geometric mean titres at day 28 statistically similar to intramuscular injection for H1N1 (1,197 vs 997; p=0.5), H3N2 (287 vs 223; p=0.4) and B strain (126 vs 94; p=0.06), with mild transient reactogenicity — tenderness in 66%, erythema in 40% and pruritus in 82% of patch recipients [12]. A measles–rubella microneedle patch trial in The Gambia enrolled 45 adults, 120 toddlers and 120 infants; in infants, 93% (52/56) seroconverted to measles and 100% (58/58) to rubella after patch administration, versus 90% (52/58) and 100% (59/59) after subcutaneous injection, with induration in 77% of toddlers and 65% of infants and discolouration in 58% and 95% respectively — all local reactions mild, with no related severe or serious adverse events [13]. A Japanese encephalitis microneedle patch trial in 39 seronegative adults delivered 0.63–1.15 µg (50–92%) of the 1.26 µg dose, achieved seroconversion in all participants, and produced localised erythema in every patch recipient [14].
These trials define the tolerability envelope a MOF patch would have to meet: erythema is expected, induration and transient discolouration are acceptable, pain is not. They also define the regulatory pathway: a vaccine-device combination with an immunogenicity endpoint.
No MOF has been tested in this setting. A PubMed search combining MOF, vaccine and transdermal or microneedle terms returned a single record, and it was not a MOF vaccine study. The immunogenicity of a framework-adjuvanted patch — which is a genuinely interesting question, since zinc and copper ions and particulate matter are all plausible adjuvants — is unstudied.
6.5 What the cargo distribution tells us about the field
The distribution of cargo classes is diagnostic. Small molecules: hundreds of papers, mature methods, real loading data. Macromolecules: dozens of papers, stabilisation data strong, transdermal pharmacokinetics absent. Nucleic acids: a handful, mostly non-therapeutic. Vaccines: effectively zero.
A field that had genuine translational intent would show the opposite gradient — starting with the hardest cargo where the clinical need is greatest, or at least progressing toward it. This field has instead deepened its coverage of the easiest cargo. That is a rational response to publication incentives and a poor predictor of clinical arrival. ## 7. Applications: Where the Evidence Is Strongest and Where It Is Thinnest
7.1 How to grade this literature
Conventional evidence hierarchies do not apply cleanly here, because there are no human studies to place at the top. A workable substitute is a four-level scheme that reflects what a clinician would need to know:
- Level A: human clinical data with a clinical or pharmacokinetic endpoint.
- Level B: animal efficacy data in more than one species, with dose–response, blinded outcome assessment and a licensed comparator.
- Level C: single-species animal efficacy against an untreated control, with mechanism data.
- Level D: in vitro only, or in vivo data reported without quantification.
On this scheme the entire MOF transdermal corpus sits at C or D. The distinctions below are therefore about relative maturity within a preclinical band, not about proximity to approval. The table summarises the assessment; the subsections explain it.
| Application | Highest evidence level observed | Quantity of work | Most credible specific finding | Principal weakness |
|---|---|---|---|---|
| Infected/chronic wounds (including diabetic) | C | Very large (~100+ records) | Consistent acceleration of closure vs untreated control; one porcine study [73] | No licensed-dressing comparator; mechanism confounded by silver/copper/zinc ions |
| Hypertrophic scar | C | Small (~6 records) | Scar elevation index reduced in rabbit ear with Zr-MOF silicone patch [50] | Single model; adsorbent mechanism poorly characterised in vivo |
| Skin cancer (melanoma, cSCC) | D–C | Moderate (~12 records) | Cytotoxicity and xenograft tumour growth delay; PDT/chemodynamic mechanisms | No lesion-level delivery data in human skin; no comparison with topical 5-FU or imiquimod |
| Acne / atopic dermatitis | C | Small (~4 records) | pH-responsive minocycline release; reduced lesion scores in mice [38][62] | No comparison with topical retinoid/steroid standard of care |
| Skin regeneration / wound-bed reconstruction | C | Moderate | Framework-loaded scaffolds support fibroblast and endothelial function | Outcome measures are histological surrogates |
| Cosmetics | D | Very small (~3 records) | Caffeine/ibuprofen patch permeation to adipose tissue ex vivo [47] | No human data; regulatory category ambiguous |
| Analgesia | D | Effectively zero | None specific to MOFs | Field absent; existing topical analgesics already established [74] |
| Anti-infective (non-wound) | D | Small | Antibacterial MICs against skin pathogens [75] | In vitro only |
7.2 Infected and diabetic wounds: the centre of gravity
A large fraction of the MOF–skin literature concerns wounds, and within that, diabetic wounds dominate. The attraction is obvious: diabetic foot ulcers are common, costly, frequently infected, and current dressings are modestly effective, so a plausible new dressing has a real clinical gap to fill. The pathophysiology — chronic inflammation, oxidative stress, impaired angiogenesis, neuropathy and biofilm — offers multiple targets for a multifunctional material, which is exactly what a MOF-plus-nanozyme-plus-microneedle construct claims to be [76][77].
The recurring result is accelerated closure versus untreated control in diabetic mice or rats. Copper-MOF hydrogels increase closure and angiogenesis [53][42]. ZIF-8 microneedles with or without copper accelerate healing and reduce scar formation [39]. UiO-66 with ultrasound reduces bacterial burden and accelerates closure [43]. Ce-UiO-66 with nerve growth factor improves healing and restores cutaneous nerve markers [69]. A bimetallic MOF microneedle array generates localised reactive oxygen species and disrupts bacterial electron transport [78]. A review of MOF–nanofibre composites states that such systems “frequently report nearly 100% wound closure within 14 days,” while acknowledging that “clinical translation remains contingent upon addressing technical and toxicological constraints,” chiefly “the lack of long term stability data and precise dosing protocols in complex wound environments” [79].
Three problems prevent these results from being clinically interpretable.
The comparator is wrong. An untreated diabetic mouse wound is not the clinical comparator; a standard-of-care dressing is. Almost no study compares against a hydrofiber, foam or negative-pressure dressing, and none compares against the dressings a wound-care service actually uses.
The active agent is ambiguous. A silver-, copper- or zinc-releasing dressing may work because of the metal ion, and ionic silver and copper dressings already exist. Demonstrating that a MOF delivers the ion more slowly than a salt is a formulation achievement; it is not a new therapeutic principle. The folic-acid-stabilised HKUST-1 paper is explicit that its advance over prior copper therapy is slower release and reduced toxicity [42] — a real but incremental gain.
The model is too forgiving. Diabetic mouse wounds heal by contraction as much as by re-epithelialisation, and splinting is used in some but not all studies to prevent contraction. Human diabetic foot ulcers have ischaemia, infection, pressure and neuropathy that no mouse model reproduces.
Against that, one study deserves specific credit: a sprayable Mg-ZIF-8 hydrogel tested in a porcine model, which reported accelerated healing with reduced iNOS-positive and increased CD163-positive macrophages, increased CD31-positive vessel formation and β3-tubulin-positive nerve regeneration [73]. Porcine skin is anatomically and immunologically closer to human skin than rodent skin, and large-animal wound data are rare in this literature. It remains a single study in a preclinical model, but it is the most translationally informative wound paper in the corpus.
7.3 Hypertrophic scar: a genuinely interesting mechanism
Scar treatment is a smaller literature with a more distinctive scientific idea. Two strands appear.
The first is adsorptive therapy. MOF-808, a zirconium framework, binds nitric oxide, cytokines and reactive oxygen species in vitro and in vivo through porosity and surface affinity rather than through a pharmacological receptor [49]. Applied as a 0.2% loaded silicone patch to rabbit ears, it reduced scar elevation and lowered TGF-β1, collagen-1 and α-SMA by 8.8%, 12% and 21.3% respectively relative to control, while the plain silicone patch group showed values higher by 21.9%, 27% and 39.9% [50]. A material that removes inflammatory mediators from a healing wound is conceptually different from a corticosteroid or a laser, and if the effect is real it would be a new mechanism class.
The second is targeted delivery to scar fibroblasts. A CD-MOF loaded with quercetin and coated with a hypertrophic-scar-fibroblast membrane, delivered by dissolvable microneedles, reduced collagen I and III and modulated Wnt/β-catenin and JAK2/STAT3 signalling, outperforming non-targeted and non-microneedle controls [33]. A calcium framework protecting collagenase achieved scar remodelling with needle dissolution in 25 minutes [63]. A hollow porphyrinic framework carrying oxygen nanobubbles used microneedle-mediated sonodynamic therapy to breach the dense fibrotic barrier [60].
The clinical context is important. Existing scar therapy — intralesional triamcinolone, silicone sheeting, pulsed dye laser, 5-FU — is modestly effective and widely used. A new approach would need to beat intralesional steroid, not untreated scar. No study in this corpus does that. Equally, the rabbit ear model is a reasonable scar model, which makes this application slightly more credible than the wound-healing literature, and the porcine and rabbit data are a small point in its favour.
7.4 Skin cancer: mechanism-rich, delivery-poor
Melanoma and cutaneous squamous cell carcinoma appear in perhaps a dozen papers, and the mechanisms are sophisticated: CD47 blockade with STING agonism [80], ferroptosis via SLC7A11/GPX4 [81], photodynamic therapy with hypoxia relief [82], photothermal plus chemodynamic therapy [61], microneedle-delivered cascade biocatalysis [83], and ZIF-8-based photodynamic therapy for cSCC [84]. A rhodium MOF in polyvinyl alcohol/polyvinylpyrrolidone nanofibres inhibited skin pathogens with MICs of 4–64 µg/mL and cancer cells with an IC50 of 19.45 µg/mL [75].
The delivery problem is severe and largely unaddressed. Melanoma is deep; a topical or microneedle approach reaches the epidermis and upper dermis at best. The papers that report tumour efficacy typically use intratumoral or intravenous administration, or a xenograft in which the delivery route is not the clinical route. For cSCC and superficial basal cell carcinoma, where topical 5-FU and imiquimod are established and effective, a MOF formulation would need a head-to-head comparison against those agents in human lesions. None exists.
The honest summary is that MOF oncology papers are pharmacology papers using skin cancer cells as a convenient cytotoxicity model. They are not transdermal delivery papers, and citing them as evidence that MOFs treat skin cancer would be a category error.
7.5 Acne and atopic dermatitis: small but well-designed
Two dermatological studies stand out for design quality.
The acne study delivered minocycline-loaded ZIF-8 from dissolvable silk microneedles into sebaceous glands, exploiting the acidic sebaceous microenvironment for release, and reported reduced lesion elevation, eschar formation and diameter scores, normalised epidermal thickness and decreased sebaceous lipid accumulation, with no detectable systemic toxicity by histology, haematology or serum biochemistry [38]. The targeting rationale — sebocyte metabolism, sebaceous gland delivery, pH-responsive release — is coherent, and the safety monitoring is broader than most.
The atopic dermatitis study used a separable microneedle patch with polydopamine-coordinated copper MOF and dictamnine, reporting an 85.1 ± 7.0% reduction in 8-OHdG, a 70.0 ± 7.8% reduction in IL-4, restored barrier integrity, a 40.1 ± 41.3% reduction in scratching frequency and a 96.6 ± 156.2% increase in time spent in the bright box of a depression assay [62]. That last confidence interval — a mean increase of 96.6% with a standard deviation of 156.2% — is worth pausing on. An effect whose standard deviation exceeds its mean by 60% indicates extreme between-animal variability, which is characteristic of small-n behavioural experiments and should temper any inference about neuropsychiatric benefit.
The clinical bar in both conditions is high: topical retinoids, benzoyl peroxide, topical corticosteroids, calcineurin inhibitors and biologics. A microneedle patch that must be applied to inflamed skin would also need to show it does not worsen irritation, which is the opposite of the usual claim.
7.6 Cosmetics and analgesia: the quiet failures
The cosmetic application is the oldest in this corpus — the 2016 MIL-100(Fe) composite patch for caffeine and ibuprofen delivered cargo to adipose tissue in an ex vivo model [47] — and has produced nothing since that has reached a consumer product. This is notable because cosmetics have the lowest regulatory bar of any application discussed here: no efficacy endpoint, no systemic safety database, no prescription. If a MOF-based cosmetic patch cannot reach market under that regime, the barrier is manufacturability and cost, not regulation.
Analgesia is effectively absent. Topical analgesics are a well-developed clinical category with established agents and evidence [74], and a review of the field notes that nanocarriers and microneedles “aim to improve efficacy but face challenges like cost and toxicity” [74]. No MOF-based analgesic patch with in vivo pharmacokinetics appears in this corpus. Given that fentanyl and lidocaine patches are the archetypal transdermal products, the absence is conspicuous and suggests the field has followed the wound-healing funding rather than the clinical need.
7.7 What the application distribution reveals
Wounds dominate because they are visually demonstrable, ethically simple, and fundable. Cancer is second because cell lines are cheap and mechanisms are publishable. Dermatology is third. Systemic indications — diabetes, pain, cardiovascular — are essentially unaddressed, even though these are where transdermal delivery would have the greatest clinical and commercial value.
This is the inverse of the commercial patch market, which is built on systemic indications. A technology that only works where the target tissue is the skin is a topical technology. That may be a perfectly good business, but it is not the “next-generation transdermal platform” the field claims to be building.
8. Biosafety, Toxicology and Pharmacokinetics: What Is Measured, and What Is Not
8.1 The measurement gap, stated precisely
A meaningful safety package for a transdermal product includes: local tolerability (irritation, sensitisation) in human skin; systemic exposure to drug, carrier and degradation products; repeat-dose toxicity; genotoxicity; and, for a nanomaterial, characterisation of the finished product’s particle size distribution, agglomeration state and stability. Regulatory guidance for nanomaterial-containing drug products addresses exactly these elements, covering quality, nonclinical and clinical considerations through development [10].
The MOF transdermal literature supplies, in the typical paper: cell viability in one or two lines at 24–48 hours; haemolysis; sometimes histology of major organs after a single application; occasionally serum biochemistry. It rarely supplies: repeat-dose data, sensitisation testing, genotoxicity, metal-ion pharmacokinetics, particle characterisation in the finished device, or mass balance.
Two reviews make this diagnosis independently. A tutorial review on MOF nanoparticle toxicity argues for systematic assessment and “safe-and-sustainable-by-design” synthesis, noting that current in vitro and in vivo studies are heterogeneous and often incomplete [36]. A second review notes that despite tens of thousands of frameworks, evaluation of their properties “in relevant biological systems remain[s] critical for their clinical translation” [29].
8.2 What the toxicology studies actually show
The most informative animal data are not about skin at all, which is itself revealing.
Cobalt. Nanoscale ZIF-67 (60 nm) impaired learning and memory in rats at a dose that produced no general toxicity, with disordered neuropeptide signalling and evidence of oxidative damage and inflammation; submicron particles (890 nm) showed better biosafety [9]. This is the single most important toxicological finding in the corpus for anyone contemplating a cobalt-containing transdermal device. It establishes that a framework can be neurotoxic at a dose that passes conventional toxicity screening, and that particle size determines the effect. The insulin microneedle study uses cobalt-doped ZIF-8 and includes an EDTA-functionalised silica chelator to capture free Co²⁺ [22] — a design response to exactly this concern, and an implicit acknowledgment of it.
Copper. HKUST-1 produced concentration-dependent cytotoxicity in HEK293 cells with mitochondrial depolarisation, apoptosis, autophagy, oxidative stress, NF-κB and NLRP3 activation and loss of zonula occludens proteins [8]. Oral MOF-74(Co) in mice produced cobalt accumulation in liver and kidney, cell-cycle arrest at 12.5 mg/L in vitro, and slight hepatic and renal inflammation on histology despite unchanged serum biochemistry [85]. The pattern across both studies is that standard biochemical panels can be normal while tissue-level injury is present.
Iron. MIL-100(Fe) showed a safe dose of 80 µg/mL in human hepatocytes and hepatoma cells with viability above 85.97% and 91.20% [48], consistent with the general view that iron carboxylates are among the better-tolerated frameworks. MIL-89 nanoparticles produced developmental toxicity in zebrafish embryos across a 1–300 µM range, with effects on viability, hatching, cardiac function and neurodevelopment [86] — an environmental toxicology study, but a reminder that “iron-based” is not synonymous with inert.
Protein corona. Human plasma proteins adsorb to framework surfaces and reduce toxicity in a copper-based pillar-layer MOF [87]. This is an important mechanistic nuance: in vivo toxicity may be lower than in vitro cytotoxicity assays predict, because the particle that reaches a cell is coated. It also means in vitro data may be poorly predictive in either direction.
8.3 Metal-ion toxicology: the arithmetic that has not been done
Consider a 1 cm² patch containing 3% w/w Cu@ZIF-8 in a 100 µm needle layer — roughly 0.3 mg of framework per cm², of which perhaps a third is copper by mass. Over 7 days, if all of it degrades, approximately 0.1 mg of copper is released into skin and potentially beyond. Even with generous assumptions about how much of that copper is absorbed, the quantity is far below the milligram-per-kilogram scale at which acute copper toxicity is usually discussed. The point is not that the number is alarming; it is that the calculation is trivial, and that no transdermal paper in this corpus performs it.
What is missing is not the arithmetic but the measurement: how much copper actually reaches the systemic circulation from a patch on intact versus damaged skin, over what time course, and whether it accumulates with repeated application. The nearest evidence is the Franz-cell work on gallium arsenide, which showed three- to six-fold higher permeation through damaged skin [55], and the general finding that skin integrity and formulation matter more than particle properties for nanoparticle penetration [6].
8.4 Deposition versus absorption: the distinction that determines everything
Section 2.3 introduced this distinction for the baseline case. In the biosafety context it becomes decisive.
If a MOF patch increases skin deposition without increasing systemic absorption, the safety question is local: irritation, sensitisation, follicular inflammation, granuloma formation, and the fate of a persistent framework in dermis. That is a manageable, testable question.
If it increases systemic absorption, the safety question is systemic: metal-ion exposure, organ accumulation, and the toxicology of the linker. That is a much larger programme.
The literature almost never distinguishes these. A typical paper reports “enhanced skin penetration” measured by fluorescence intensity in skin sections, which is deposition. The 5-FU/ZIF-8 study is the exception that proves the rule: it measured both and found opposite directions [5]. Any clinician reading a MOF transdermal paper should look for the receptor-compartment data. If there is none, the paper has not measured absorption.
8.5 Irritation and sensitisation: the missing human data
Cutaneous reactions are the most common adverse effect of existing transdermal systems [2]. For a new platform, the relevant testing framework is the one used for medical devices and topical products — irritation and sensitisation studies in human volunteers, typically under occlusion, with repeat application.
No MOF transdermal study in this corpus reports human irritation or sensitisation testing. The device literature offers a comparator: dissolving microneedle patches produced no pain or swelling and only mild erythema resolving within seven days in 15 volunteers [88], and no severe adverse reactions in 20 volunteers in an earlier immunization study [89]. Those are the numbers a MOF patch would need to match. They have not been generated.
8.6 Pharmacokinetics: what a clinician would need
For a systemic transdermal product, the required data are: steady-state plasma concentration, area under the curve relative to an intravenous or oral reference, time to steady state, inter-individual variability, and the effect of application site, skin temperature and integrity. In vitro–in vivo correlation is the regulatory framework for predicting these from bench data, and it is a recognised discipline with its own modelling approaches and failure modes [28].
For a local dermatological product, the required data are different: concentration in the target compartment (epidermis, dermis, sebaceous gland), residence time, and systemic exposure as a safety parameter rather than an efficacy one.
The MOF literature supplies neither. In the entire corpus reviewed here, we found no plasma concentration–time curve for a MOF-delivered drug, no bioavailability estimate, and no human pharmacokinetic data of any kind. The only human data for any MOF-containing microneedle device concern a diagnostic sensor measuring NGAL in interstitial fluid, where MOF encapsulation preserved antibody activity through four weeks of thermal cycling to 50 °C and measurements correlated with blood NGAL in mice and human subjects [7]. That is a genuinely impressive piece of engineering. It is also not a therapeutic pharmacokinetic dataset.
8.7 The framework as a persistent object
A final safety consideration applies specifically to stable frameworks. UiO-66, MOF-808 and SU-102 are designed not to degrade quickly. If a microneedle delivers such particles into dermis, they remain there. What happens to a persistent crystalline coordination polymer in dermis over months is unknown: whether it is phagocytosed, whether it provokes foreign-body reaction, whether it translocates to draining lymph nodes.
The adsorbent scar study is the clearest case in point. Its therapeutic mechanism depends on the framework remaining intact to bind mediators [49]. A material that must persist to work must be shown to be safe while persisting. No study in this corpus addresses that.
9. Manufacturing, Quality, Sterilisation and Scale-Up
9.1 The gap between a paper and a product
A transdermal patch is a manufactured article with a specification: assay, uniformity of content, adhesion, release rate, microbial limits, extractables and leachables, and shelf life. A MOF microneedle patch adds a second layer of complexity: a nanomaterial with its own critical quality attributes (particle size, crystallinity, surface area, residual solvent, metal speciation) embedded in a dissolving polymer matrix with its own critical attributes (mechanical strength, dissolution time, needle geometry).
No paper in this corpus presents a specification, a stability protocol, or a lot-release strategy. That is not a criticism of individual studies — it is outside their scope — but it means the manufacturing literature must be assessed separately, and it is thin.
9.2 Synthesis at scale
Framework synthesis is the first bottleneck. Most biomedical MOFs are made solvothermally in dimethylformamide or dimethyl sulfoxide at elevated temperature, which is unacceptable at scale for solvent-residue reasons. ZIF-8 is the exception: it assembles in water or methanol at room temperature, which is a large part of why it dominates.
The scale-up literature that exists is encouraging but narrow. A microwave-assisted route to UiO-66 optimised across a wide parameter space achieved scale-up while maintaining efficiency, with no significant differences in morphology, crystal size or defects compared with conventional heating [37]. That is a genuine manufacturing contribution: it shows a benchmark framework can be made reproducibly at larger scale with a faster process.
What it does not show is pharmaceutical quality. Residual linker, unreacted metal salt, occluded solvent, amorphous by-product and particle-size distribution all require control, and none is reported.
9.3 Sterilisation
A product applied to broken skin or a microneedle that breaches the stratum corneum must be sterile or terminally sterilised. The options — moist heat, gamma irradiation, ethylene oxide, electron beam — each interact with the material. Gamma irradiation is known to induce structural degradation in ZIF-8, with electron-beam studies quantifying dose tolerance [90]. Ethylene oxide requires aeration and leaves residues. Moist heat would hydrolyse acid-labile frameworks.
No transdermal MOF paper in this corpus reports sterilisation validation. This is a concrete, solvable, and completely unaddressed engineering problem, and it is the kind of issue that kills programmes late.
9.4 Stability and shelf life
ZIF-8 is hydrolytically labile; CD-MOFs are described as having “aqueous instability” as a principal limitation [54]; HKUST-1 “disintegrated in protein solution” unless protected by a hydrogel [53]. A product that degrades in the presence of water and protein must be protected by packaging and formulation, and its shelf life must be demonstrated.
The one stability success in the corpus is instructive because it uses the framework’s protective capacity rather than its delivery capacity: a ZIF-8 shell grown in situ on antibody-functionalised microneedles preserved antibody bioactivity after four weeks of thermal cycling to 50 °C and maintained analytical performance after prolonged unrefrigerated shipment [7]. That is a cold-chain-elimination result with real public-health value. It is also a reminder that the most clinically valuable MOF application found in this entire review is a diagnostic stability technology, not a therapeutic delivery system.
9.5 Device manufacturing economics
The microneedle field, which is a decade ahead of the MOF field in manufacturing, provides the relevant cost model. A 2026 economic analysis reports that microneedle technology has generated more than 4,000 patents but commercial translation remains limited to 13 products, predominantly cosmetic; that patches achieve cost-effectiveness at price points below US$30 per dose; that operational savings are US$0.24–0.61 per dose in vaccination scenarios, driven mainly by labour; and that establishing a production facility requires US$5–20 million in capital [91].
Those numbers define the hurdle. A MOF patch would carry the same device costs plus nanomaterial manufacturing, characterisation and quality-control costs. For a vaccine priced at a few dollars per dose, the economics do not close without a value proposition beyond convenience. For a chronic wound dressing reimbursed at tens of dollars, they might — which is consistent with the field’s concentration on wounds.
9.6 Batch reproducibility and the “same material” problem
A recurring methodological weakness is that “ZIF-8” in one paper is not the same object as “ZIF-8” in another. Particle size, morphology, defect density, surface coating and residual solvent all vary with synthesis conditions, and a systematic review of factors affecting MOF characteristics found that additives, ligand choice and synthesis parameters substantially alter size and structure [92].
For a regulator, this is the central quality question: can the material be specified precisely enough that batch-to-batch variation does not change safety or efficacy? For a reader of the literature, it means that a result obtained with one preparation may not replicate with a nominally identical one — a problem that is well recognised in nanomedicine generally, where the generic “sameness” paradigm does not apply and a “similarity” approach is required instead [93].
9.7 What a credible development programme would look like
If a sponsor were serious, the sequence would be: define a lead candidate with a specified synthesis and purification; demonstrate scalable manufacture with in-process controls; validate sterilisation; establish shelf life under ICH conditions; characterise the finished device (needle geometry, mechanical strength, dissolution, residual moisture); perform formal biocompatibility testing; then enter first-in-human studies with local tolerability and pharmacokinetics as endpoints.
None of these steps requires a scientific breakthrough. All of them require money and time, and none of them is rewarded by the current publication system. That is the structural reason the field has not advanced: the incentive to do the boring work is absent, not the capability. ## 10. Clinical and Regulatory Status: What Has Reached Humans, and How a Product Would Be Classified
10.1 The registry search, reported in full
The central factual claim of this review is that no MOF-based transdermal therapeutic has entered clinical testing. That claim should be verifiable, so the searches are reported here in full, including the failures.
ClinicalTrials.gov v2 API searches were run on 2026-10-30 with query.term set to each term below and countTotal=true. The v2 query engine performs a text match across study fields, so counts include false positives; the relevant studies were inspected individually.
| Search term | Total records returned | Relevant to MOF transdermal delivery |
|---|---|---|
metal-organic framework |
5 | 0 |
MOF transdermal |
2 | 0 |
metal organic framework skin |
1 | 0 |
ZIF-8 |
1 | 0 |
UiO-66 |
0 | 0 |
HKUST-1 |
0 | 0 |
MOF nanoparticle |
1 | 0 |
metal-organic frameworks drug delivery |
0 | 0 |
microneedle |
165 | 165 (device, not MOF) |
microneedle patch vaccine |
11 | 11 (device, not MOF) |
dissolving microneedle |
3 | 3 (device, not MOF) |
The five records returned for “metal-organic framework” were: an extension study in interstitial lung disease (NCT06644144), a Colgate toothpaste trial (NCT00926029), a tegoprazan-based H. pylori eradication trial (NCT06977841), an olaparib/cediranib small-cell lung cancer trial (NCT02899728), and a study of decision-making deficits in traumatic brain injury (NCT02169310). None involves a metal–organic framework. The two hits for “MOF transdermal” were a nicotine patch trial in COVID-19 (NCT04598594) and an endoscopic nitroglycerin trial for post-ERCP pancreatitis (NCT07083063) — both matched on unrelated tokens. The single hits for “ZIF-8” and “MOF nanoparticle” were the toothpaste trial again and an exosome proteomics study in sepsis (NCT03267160).
The count of registered clinical trials of MOF-based transdermal drug delivery is zero. This is not a case of a small number of early studies; it is an empty set. The same is true of MOF-based therapeutics by any route in these searches, which is consistent with the broader observation that MOF biomedical research has not yet produced a marketed product.
10.2 The one MOF device with human data
The exception that should be stated precisely is a diagnostic device. A ZIF-8-encapsulated microneedle patch measured neutrophil gelatinase-associated lipocalin in dermal interstitial fluid, preserved antibody activity through four weeks of thermal cycling to 50 °C, and in a mouse model and human subjects produced measurements that closely mirrored blood NGAL, preceded conventional renal function markers, and correlated with histopathological injury severity [7].
This is a real translational achievement: a MOF-containing device, applied to human skin, producing analytically valid results. It is also a biosensor, not a therapy. It delivers nothing, so it does not face the pharmacological and toxicological questions that a therapeutic patch faces. Its existence demonstrates that the materials and manufacturing problems are surmountable; it does not demonstrate that therapeutic delivery works.
A second device-adjacent example is a MOF-on-MOF electrochemical microneedle array for uric acid in interstitial fluid, with a limit of detection of 20 nM and recovery of 98.6–100.4% in ex vivo rat skin [94]. Again: sensing, not delivery.
10.3 The realistic benchmark: dissolving microneedle patches in humans
Because the MOF field has no human data, the relevant benchmark is the technology it would be layered onto. Dissolving microneedle patches have now been through randomised clinical trials with immunogenicity endpoints, and the results define what “ready for humans” looks like.
| Trial | Design | Key results | Tolerability |
|---|---|---|---|
| Influenza TIV-MNP 2015 [12] | Phase 1, randomised, partly blinded, placebo-controlled, n=100 | Day-28 GMTs similar to IM: H1N1 1,197 vs 997 (p=0.5); H3N2 287 vs 223 (p=0.4); B 126 vs 94 (p=0.06) | Tenderness 66%, erythema 40%, pruritus 82% with patch; no related SAEs |
| Measles–rubella, The Gambia [13] | Phase 1/2, double-blind, double-dummy, active-controlled, age de-escalation; 45 adults, 120 toddlers, 120 infants | Infant seroconversion: measles 93% (52/56) vs 90% (52/58) SC; rubella 100% both | Induration 77% toddlers, 65% infants; discolouration 58% and 95%; all mild; no related severe AEs |
| Japanese encephalitis MNA-J [14] | Phase 1, randomised, partly blinded, active-controlled, n=39 (13/group) | Delivered 0.63–1.15 µg (50–92%) of 1.26 µg; all seroconverted; mean log10 titres 2.55 (25%), 2.04 (10%), 2.08 (SC) | Localised erythema in all patch recipients; no SAEs |
| Placebo tolerability [88] | 15 volunteers | — | No pain or swelling; mild erythema resolving within 7 days; self-administration feasible |
| Early immunisation device [89] | 20 volunteers | Antibody titre increase comparable to prior device | No severe adverse reactions |
The pattern is consistent and clinically legible: immunogenicity approximately equivalent to injection, local reactions that are frequent but mild and self-limiting, and no systemic safety signal in several hundred participants. Reviews of the clinical microneedle landscape describe a field with multiple products approved for cosmetic and therapeutic use and numerous trials ongoing [95][96].
This is the bar. A MOF microneedle patch would need to demonstrate, at minimum, comparable local tolerability, plus framework-specific safety data that no microneedle-only product requires. Nothing in the current literature suggests that bar has been approached.
10.4 How a MOF-eluting patch would be regulated in the United States
A drug-loaded microneedle patch is a combination product. Under 21 CFR 3.2(e), a combination product includes “a product comprised of two or more regulated components, i.e., drug/device, biologic/device, drug/biologic, or drug/device/biologic, that are physically, chemically, or otherwise combined or mixed and produced as a single entity” [11]. A patch containing a drug in a microneedle array is exactly that: a drug constituent part and a device constituent part produced as a single entity.
Two consequences follow. First, the product needs a lead centre assignment — typically CDER for a drug-primary product, with CDRH consultation on the device constituent. Second, the device constituent must meet the applicable device requirements (design controls, biocompatibility, sterility) in addition to the drug requirements.
If the framework itself is a nanomaterial — and a 50–500 nm coordination polymer is — then FDA’s nanotechnology guidance applies. The 2014 guidance defines materials in the nanoscale range as those “with at least one dimension in the size range of approximately 1 nanometer (nm) to 100 nm,” while noting that products may exhibit nanoscale properties at dimensions outside that range [97]. That caveat matters here: MOF particles are frequently larger than 100 nm yet exhibit the high surface-area-to-volume behaviour that motivates nanomaterial scrutiny.
The 2022 final guidance, “Drug Products, Including Biological Products, That Contain Nanomaterials,” applies to human drug products “in which a nanomaterial is present in the finished dosage form,” and addresses “quality, nonclinical, and clinical studies” as they relate to such products throughout development and production, including specific considerations for abbreviated pathways [10]. The practical implication is that a sponsor would need to characterise the nanomaterial in the finished product — not just the bulk powder — and to address nanomaterial-specific quality, nonclinical and clinical questions rather than relying on conventional generic approaches.
For a nanomedicine more broadly, the regulatory literature is clear that the generic “sameness” paradigm does not transfer: because nanomedicines are heterogeneous and cannot be fully characterised physicochemically, a similarity approach is required for follow-on versions [93]. A framework whose particle size distribution, defect density and surface coating vary between batches is precisely the kind of product this concern targets.
10.5 The European framing
In the EU, a medicine that requires a device to be administered is assessed as a medicinal product through the centralised procedure, with the device component subject to conformity assessment; where a medical device incorporates an ancillary medicinal substance, the notified body must obtain a scientific opinion from EMA or a national competent authority before issuing a CE certificate [98]. EMA also has a specific regulatory track for “medical devices made of substances that are systemically absorbed,” which is the category most likely to capture an absorbable framework-based system [98].
The classification question — medicinal product versus medical device — is therefore decided by principal mode of action, not by physical form. A patch whose primary action is pharmacological (delivering a drug) is a medicinal product; a patch whose primary action is physical (barrier, adsorption, mechanical) could be a device. This is directly relevant to the adsorptive scar patch discussed in Section 7.3, whose mechanism is adsorption of inflammatory mediators rather than pharmacology [49]. That product might plausibly be regulated as a device in Europe and as a device-drug combination in the United States — a classification difference with substantial consequences for the evidence required.
The EU primary texts (Regulation 2017/745 and the EMA nanomedicine reflection paper) could not be retrieved in this environment; the description above rests on EMA’s own regulatory overview page. This limitation is recorded in Section 14.
10.6 What the regulatory position means practically
Three conclusions follow.
First, there is no regulatory shortcut. A MOF transdermal product would need the full package: quality, nonclinical, clinical, plus device requirements. The framework adds characterisation burden rather than removing it.
Second, the classification is not obvious, and that is a risk. A sponsor uncertain whether the product is a drug, a device or a combination faces additional early regulatory interactions and possible reclassification later. The adsorbent-scar case shows this is not hypothetical.
Third, the first successful product is likely to be a device-like topical, not a systemic transdermal. A wound dressing or scar patch with a local mechanism, a bounded duration and a device pathway is a far easier first approval than a systemic transdermal patch that must demonstrate bioavailability and chronic systemic metal exposure.
11. Artificial Intelligence in MOF Design: Demonstrated Versus Aspirational
11.1 What the seed claims
The seed review states that artificial intelligence “may accelerate MOF design, predict structure–property relationships, optimize stimuli-responsive release, and support the development of personalized transdermal therapies” [3]. Every verb in that sentence is hedged, and correctly so — it is a claim about possibility, not about accomplishment, and the seed’s abstract reports no AI results of its own.
The question for this review is narrower: what has AI actually done for MOFs, and has any of it touched transdermal formulation?
11.2 Demonstrated: property prediction and literature mining
The demonstrated capabilities fall into three groups, all of them upstream of formulation.
Property prediction from structure. A machine-learning model predicted MOF guest accessibility — whether a framework’s pores can accommodate a guest molecule — with 80.5% certainty based solely on the identity of the metal and linker chosen by the experimentalist, by decomposing experimental crystal structures from the Cambridge Structural Database and learning the relationship between component chemistry and porosity [99]. This is a genuine, useful, and modest capability: it ranks candidate metal–linker combinations before synthesis, saving bench time.
Text mining and data extraction. A text-mining algorithm extracted surface area and pore volume from MOF manuscript HTML files with 90% and 88.8% accuracy on a 200-MOF training set, falling to 73.2% and 85.1% on a randomly selected test set [100]. A larger effort data-mined 46,701 MOFs from 28,565 papers to extract synthesis parameters with an average F1 score of 90.3%, then used positive-unlabelled learning to predict whether a given synthesis would succeed, correctly predicting 83.1% of synthesised materials in the test set [101]. These are the least glamorous and most immediately useful AI results in the field: they turn an unruly literature into a database.
Multimodal prediction from synthesis-accessible inputs. A 2025 model used only the powder X-ray diffraction pattern and the synthesis chemicals — information available as soon as a material is made — to predict pore structure, chemistry-dependent and quantum-chemical properties, with self-supervised pretraining on database crystal structures to compensate for small experimental datasets [30]. The authors frame the problem in terms that will be familiar to anyone who has watched a materials field over-promise: over 5,000 new MOFs are reported each year, and “linking these new materials to their best applications remains a significant challenge” [30].
11.3 Demonstrated: language models as research infrastructure
A 2025 perspective on AI paradigms for MOF research is candid about where large language models currently add value: “the most common applications in metal-organic framework (MOF) research include automating literature reviews and data extraction to accelerate the material discovery process” [102]. That is a fair summary. LLMs are being used to read papers, extract numbers and draft summaries — useful, unglamorous, and a long way from designing a transdermal patch.
A broader review of machine learning for MOF discovery covers regression, classification, clustering, deep learning and reinforcement learning, and is similarly focused on gas storage, separation and catalysis rather than biomedical formulation [103].
11.4 Aspirational: AI for transdermal formulation
The transdermal AI literature exists, but it is not MOF-specific. A 2025 review of AI-powered transdermal systems asserts that machine-learning models “predict drug entrapment with 93.0% accuracy,” forecast release kinetics, skin penetration and stability, identify permeation enhancers by virtual screening, optimise microneedle geometry and enable adaptive dosing through biosensors [104]. Those are the claims of a review, not primary results, and the accuracy figure is reported without the dataset, endpoint definition or validation design that would let a reader judge it.
No paper in this corpus applies machine learning to MOF selection or formulation for transdermal delivery. The intersection is empty. The seed’s suggestion that AI could “optimize stimuli-responsive release” and “support personalized transdermal therapy” is therefore a reasonable extrapolation from adjacent capabilities and an unsupported claim about this field.
11.5 What AI could realistically contribute, and what it cannot
The realistic contributions are worth stating because they are genuinely valuable and should not be dismissed along with the hype.
Candidate prioritisation. Given a cargo and a target, a model could rank frameworks by predicted loading, predicted degradation kinetics at skin pH, and predicted metal-ion release — all properties that could be learned from existing data if the data were curated. The guest-accessibility model is a proof of principle [99].
Formulation screening. Predicting which polymer–framework–drug combination will produce a mechanically adequate, stable microneedle is a multivariate optimisation problem well suited to ML, and the microneedle field already has the datasets.
Literature synthesis. Extracting comparable loading, release and toxicity values from thousands of papers would produce the systematic evidence base this field currently lacks. The text-mining results suggest this is feasible [101][100].
What AI cannot do is generate the missing data. No model can predict human skin permeation of a framework from a literature that contains no human skin permeation data. The binding constraint is experimental, not computational, and the field’s enthusiasm for AI risks becoming a way of avoiding the expensive experiments rather than a way of designing them better.
12. Evidence-Quality Appraisal: Quantity, Quality and the Translation Gap
12.1 The quantitative picture
To characterise the literature rather than merely describe it, we screened 201 PubMed records at the intersection of MOF terms and skin/transdermal terms, retrieved by 64 distinct PubMed queries logged during this review (part of 80 logged queries across three databases). The distribution is striking.
| Metric | Value |
|---|---|
| MOF × skin/transdermal records screened by title | 201 |
| Published 2024 or later | 153 (76%) |
| Published 2026 or later | 67 (33%) |
| Abstracts read in full for this review (MOF × skin subset) | 48 |
| Of those, reporting animal models | 22 (46%) |
| Of those, mentioning patients or human subjects at all | 2 (4%) |
| Of those, reporting a clinical study of a MOF therapeutic | 0 |
| Registered clinical trials of MOF transdermal delivery | 0 |
The publication curve is the single most informative statistic in this review. Three-quarters of the entire corpus appeared in the last two years. A field producing 64 papers in 2026 on a technology with no clinical candidate is not converging on translation; it is expanding a preclinical design space. Both can be legitimate, but they are different activities, and the growth rate should not be mistaken for progress toward a product.
The 46% figure for animal work should be read as an upper bound on rigour, not a lower bound: a paper that mentions a mouse model may still be single-species, single-dose, unblinded and uncontrolled beyond an untreated comparison. The 4% figure for any mention of human subjects is more telling. In a corpus of 48 papers about a technology intended for human skin, two mention humans, and neither reports a therapeutic study.
12.2 Quality dimensions that cannot be counted
Several quality problems recur across the corpus and resist tabulation.
Absence of dose–response. Nearly every efficacy study compares one dose of the formulation against untreated control. Without a dose–response curve, the claim that the framework improves delivery is indistinguishable from the claim that more drug was applied.
Absence of active comparators. The clinical standard of care — a silver dressing, a hydrofiber, a topical corticosteroid, topical 5-FU — is almost never the comparator. Where a free-drug control exists, it is usually the same drug in solution at the same nominal dose, which does not control for the formulation’s own effects.
Absence of blinding and randomisation reporting. Animal experiments in this literature rarely state whether outcome assessment was blinded or how animals were allocated. Wound-area measurement from photographs is susceptible to substantial observer bias, and histological scoring more so.
Small n and wide variance. The atopic dermatitis study reported a 96.6% mean increase in a behavioural outcome with a standard deviation of 156.2% [62]. Whatever the biological reality, an effect size with that dispersion cannot support a confident inference, and the paper’s conclusion — that the system “alleviated depression-like behavior” — outruns its statistics.
Surrogate endpoints. Wound closure percentage, scar elevation index, fluorescence intensity and histology scores are surrogates. None is a patient-important outcome, and the correlation between them and clinical benefit in humans is assumed rather than demonstrated.
Mechanism claims from correlation. RNA sequencing after treatment showing NF-κB pathway changes [69] identifies a pathway that moved, not the mechanism of the material. Many papers present pathway analysis as mechanistic proof when it is hypothesis generation.
12.3 The one systematic review in the field
A systematic review of factors affecting MOF characteristics for biomedical applications searched four databases and included 20 eligible studies [92]. Twenty. For a field with tens of thousands of synthesised frameworks and hundreds of biomedical papers, a systematic review that could identify only 20 eligible studies reflects the absence of standardised reporting rather than a narrow question. The review’s finding — that additives, ligands and synthesis parameters substantially affect size and structure — is unsurprising but important, because it means the “same” framework can behave differently across papers.
There is no systematic review of MOF transdermal delivery, no meta-analysis, and no prospect of one: the heterogeneity of frameworks, cargos, models and outcomes makes pooling impossible. The seed review is a narrative review, and the field’s most cited transdermal syntheses are narrative reviews as well [4][77][105].
12.4 Citation momentum versus clinical momentum
Forward-citation counts illustrate the gap between scientific attention and translational progress. As of the date of preparation, Europe PMC recorded 144 citations for the photothermal MOF microneedle wound paper [41] and 58 for the glucose-responsive insulin microneedle paper [22] [106]. Both are well-cited, well-designed studies. Neither has produced a follow-on clinical programme, a licensed product, or a human pharmacokinetic dataset.
High citation counts in this field measure interest among materials scientists, not proximity to patients. That is not a criticism of the science; it is a warning about how the science is described in reviews and press releases.
12.5 The translation gap, quantified as a pipeline
A useful way to express the gap is as a funnel:
- Thousands of MOF structures synthesised for biomedical applications.
- Hundreds evaluated in a skin or wound model.
- Tens evaluated in a second species or with formal toxicology.
- Zero entered a clinical trial for transdermal delivery.
- Zero marketed.
The drop between the third and fourth level is the one that matters, and it is not caused by a single missing experiment. It is caused by the cumulative absence of the unglamorous package described in Section 9.7: specification, scalability, sterilisation, shelf life, biocompatibility, pharmacokinetics.
This is the same gap that separates most nanomedicine from products, and it has been described for two decades. What is specific to MOFs is the metal. A lipid nanoparticle’s degradation products are lipids; a framework’s are a metal salt and an organic acid, and the metal is the part that regulators and toxicologists will want quantified.
12.6 What would raise the evidence level
Concrete, achievable steps would move the field from C/D to B on the scale in Section 7.1:
- A dose-ranging efficacy study with three doses, a licensed comparator and blinded outcome assessment in a large-animal wound model.
- A mass-balance pharmacokinetic study after transdermal application, reporting skin, lymph node and plasma concentrations of drug, metal and linker over 28 days.
- A repeat-dose local tolerability study in human volunteers, using the microneedle-patch methodology already established [88].
- A sterilisation and stability validation for a defined finished device.
- A negative-control device — identical microneedle, no framework — in every efficacy study.
None of these requires new chemistry. Each would substantially increase confidence. The fact that almost none has been done is the most important finding of this appraisal.
13. Bottom Line for Clinicians
13.1 What this changes today
Nothing. There is no MOF-based transdermal product, no clinical trial, no human pharmacokinetic dataset, and no regulatory submission in the public record. A clinician will not encounter one in practice within the likely shelf life of this review, and should be sceptical of any claim otherwise.
13.2 What is genuinely interesting, and why
Three things in this literature are worth a clinician’s attention, none of them ready for use.
The loading capacity is real. Framework loading of 30–40% w/w for small molecules, and approximately 40 wt% for insulin, exceeds what conventional nanocarriers achieve [31][32]. If a microneedle can carry more drug per needle, patches can be smaller and cheaper. That is a manufacturing advantage with plausible clinical consequences.
The stimuli-responsiveness is chemically elegant. A framework that degrades specifically in the acidic environment of an infected wound or a tumour, releasing its cargo where it is needed, is a more sophisticated trigger than a polymer that erodes on a fixed timetable [21][22]. Whether the specificity is sufficient in the messy pH range of a real wound is untested.
The barrier problem has been correctly diagnosed. The best papers in this field are explicit that MOFs do not cross skin and that the microneedle is doing the work [4]. That honesty is a precondition for progress, and it distinguishes the better reviews from the promotional ones.
13.3 What would constitute real progress
A clinician should look for the following, in roughly this order, and should discount any claim that arrives without them.
Human skin, human subjects. In vitro permeation through human epidermis with mass balance, then a phase 1 study with local tolerability as the primary endpoint. The microneedle field has shown this is achievable and what the results look like [88][12]. The absence of any such study for MOFs is the field’s defining deficiency.
A pharmacokinetic curve. For any systemic claim, a plasma concentration–time profile with a bioavailability estimate against a reference route. For any local claim, target-tissue concentrations with systemic exposure as a safety parameter. No such data exist for MOF transdermal delivery.
Metal mass balance. How much zinc, copper or zirconium leaves the patch, how much stays in skin, how much reaches plasma and organs, and what happens on repeat application. The tools exist — Franz cells with elemental analysis, ICP-MS, IVIS imaging — and the gallium arsenide work shows what a careful permeation study looks like [55].
A licensed comparator. A wound dressing that a wound-care service actually uses, or topical 5-FU for a skin lesion, or intralesional steroid for a scar. Not an untreated control.
Reproducible manufacture. A specified synthesis, a defined finished-device specification, validated sterilisation and a shelf life. The microwave UiO-66 scale-up is the kind of work needed, extended to pharmaceutical quality [37].
Regulatory clarity. An early classification decision — drug, device or combination — and a characterisation package aligned with nanomaterial guidance [10][11].
13.4 The failure modes a clinician should watch for
If a MOF transdermal product does reach trials, the likely problems are predictable and worth naming in advance.
Local irritation and sensitisation. The framework, the linker and the metal are all potential sensitisers. Existing patches already cause contact dermatitis at clinically meaningful rates [2], and microneedles add mechanical trauma.
Metal accumulation with chronic use. A cosmetic or analgesic patch used daily for years delivers a continuous metal load. Cobalt’s neurotoxicity at non-general-toxic doses in rats is the cautionary example [9].
Particulate persistence. Stable frameworks left in dermis or follicles may provoke foreign-body reactions or accumulate in draining lymph nodes. No chronic data exist.
Follicular pathology. Follicular delivery is a feature for acne and a risk for folliculitis and granuloma in other skin types.
Antimicrobial resistance pressure. Silver- and copper-releasing dressings expose wound flora to sub-lethal metal concentrations. Whether framework-controlled release reduces or increases resistance selection is unknown and unstudied.
13.5 The honest framing
Transdermal delivery is a mature, useful, and genuinely limited route. The limitations are physical: the stratum corneum is an excellent barrier, and the molecules that cross it are a small, well-characterised set. Metal–organic frameworks are an interesting class of materials with genuinely unusual loading and release properties. The combination is a reasonable research hypothesis.
It is not a therapeutic advance, and the language used to describe it should reflect that. A paper showing that a MOF microneedle accelerates wound closure in diabetic mice has shown that a material can be made and that a mouse healed. It has not shown that a patient would benefit, that the material is safe, that the product can be manufactured, or that a regulator would approve it. Those are four separate, expensive, and currently unattempted projects.
The most useful thing a clinician can take from this literature is a calibrated expectation: watch for human pharmacokinetic data and repeat-dose tolerability, treat everything before those as formulation science, and be unmoved by wound-closure photographs until there is a comparator, a dose–response and a blinding statement.
14. Limitations
Evidence sourced from abstracts, not full texts. For all 101 cited PubMed records, the abstract was retrieved and read; full texts were not downloaded. Numbers quoted are those reported in abstracts, and claims that depend on details available only in full text — exact skin models, randomisation methods, vehicle composition — could not be verified. Where a figure is abstract-reported, this review says so.
The seed review is a narrative review with no quantitative abstract. It was cited only for its qualitative framing: the stratum corneum and target-accumulation problems, the claimed MOF advantages, the named application areas, the translational barriers, and the AI speculation [3]. No number in this review is attributed to the seed. It is not a systematic review and reports no effect sizes in its abstract, and its evidence base is preclinical.
Purposive, not systematic, sampling. The 201 records at the MOF–skin intersection were identified by 64 PubMed queries and screened by title; 48 of them were read in full-abstract form. The selection was purposive — weighted toward microneedle hybrids, framework classes, toxicology and clinical-adjacent work — and the quantitative statements in Section 12 describe that sample, not a random sample of the field. A formal systematic review with dual screening would produce different counts.
ClinicalTrials.gov searching is text-based and imperfect. The v2 API matches terms across study fields and returns false positives, as documented in Section 10.1. A trial registered under an unusual title using none of the searched terms could have been missed. WHO ICTRP and national registries were not searched, so the “zero trials” finding is robust for ClinicalTrials.gov but not exhaustive across all registries.
Regulatory sources are asymmetric. The FDA position rests on primary documents retrieved and read: the 2014 nanotechnology guidance, the 2022 Federal Register notice of the nanomaterials drug products guidance, and 21 CFR 3.2 [97][10][11]. The EU position rests on EMA’s medical devices overview page [98]; Regulation 2017/745 and EMA’s nanomedicine reflection paper could not be retrieved in this environment because EUR-Lex and the relevant EMA pages were unreachable. EU statements are therefore at a general level and should not be relied on for a specific classification decision.
The 2022 nanomaterial guidance was read as a notice, not as the guidance document. The Federal Register notice describes the guidance’s scope and content; the guidance PDF itself was not retrieved. Statements about what the guidance addresses are drawn from the notice’s summary.
Contradictory or ambiguous terminology in the source literature. “Transdermal” is used loosely in this field to mean topical, dermal, follicular, microneedle-mediated, and occasionally transdural [51]. Every effort was made to interpret each paper’s route from its methods rather than its title, but misclassification is possible.
No assessment of publication bias or file-drawer effects. Negative results in this field are unlikely to be published, and the true effect size of MOF-based wound healing is therefore probably smaller than the literature suggests. No funnel plot or registry of preclinical protocols was available.
Citation counts are point-in-time. The Europe PMC counts in Section 12.4 were retrieved on the date of preparation and will change [106].
Single-reviewer synthesis. Searches, screening, extraction and drafting were performed by one process without a second independent screener. Errors of omission are more likely than in a dual-reviewer systematic review.
Not clinical guidance. Nothing in this review should be used to select, recommend or justify a therapy. It is an assessment of a research field, and its principal conclusion is that the field has not yet generated evidence that bears on clinical practice.
References
- Ghaferi M, Alavi SE, Phan K, Maibach H, Mohammed Y. Transdermal Drug Delivery Systems (TDDS): Recent Advances and Failure Modes. Mol Pharm. 2024;21(11):5373-5391. doi:10.1021/acs.molpharmaceut.4c00211. PMID: 39365887.
- Bershow A, Warshaw E. Cutaneous reactions to transdermal therapeutic systems. Dermatitis. 2011;22(4):193-203. PMID: 21781635.
- Ma D, Guo S, Liu R, Feng Y, Nezamzadeh-Ejhieh A, Lu L, et al. Advances in MOF-Based Transdermal Drug Delivery Systems: Mechanisms, Applications, and Future Prospects. Top Curr Chem (Cham). 2026;384(4). doi:10.1007/s41061-026-00566-8. PMID: 42763813.
- Jing Y, Liu X, Zhu Y, Wu L, Nong W. Metal-organic framework microneedles for precision transdermal drug delivery: design strategy and therapeutic potential. Nanoscale. 2025;17(10):5571-5604. doi:10.1039/d4nr03898c. PMID: 39918280.
- Wang Q, Li M, Sun X, Chen N, Yao S, Feng X, et al. ZIF-8 integrated with polydopamine coating as a novel nano-platform for skin-specific drug delivery. J Mater Chem B. 2023;11(8):1782-1797. doi:10.1039/d2tb02361j. PMID: 36727421.
- Gimeno-Benito I, Giusti A, Dekkers S, Haase A, Janer G. A review to support the derivation of a worst-case dermal penetration value for nanoparticles. Regul Toxicol Pharmacol. 2021;119:104836. doi:10.1016/j.yrtph.2020.104836. PMID: 33249100.
- Wang Y, Liu Y, Debnath A, Li C, Wang Z, Kondepati G, et al. Metal-Organic Framework-Preserved Thermostable Microneedle Patch for Minimally Invasive Detection and Monitoring of Kidney Dysfunction. Adv Mater. 2026;38(55):e74423. doi:10.1002/adma.74423. PMID: 42554422.
- Chen YC, Andrew Lin KY, Chen KF, Jiang XY, Lin CH. In vitro renal toxicity evaluation of copper-based metal-organic framework HKUST-1 on human embryonic kidney cells. Environ Pollut. 2021;273:116528. doi:10.1016/j.envpol.2021.116528. PMID: 33486253.
- Deng S, Yan X, Xiong P, Li G, Ku T, Liu N, et al. Nanoscale cobalt-based metal-organic framework impairs learning and memory ability without noticeable general toxicity: First in vivo evidence. Sci Total Environ. 2021;771:145063. doi:10.1016/j.scitotenv.2021.145063. PMID: 33736171.
- US Food and Drug Administration. Drug products, including biological products, that contain nanomaterials: guidance for industry; notice of availability. Fed Regist. 2022;87(78):24169-70. Docket FDA-2017-D-0759. https://www.federalregister.gov/documents/2022/04/22/2022-08572/drug-products-including-biological-products-that-contain-nanomaterials-guidance-for-industry
- 21 CFR 3.2(e). Combination product: definition. Code of Federal Regulations, Title 21, Part 3. Current as of 2026. https://www.ecfr.gov/api/renderer/v1/content/enhanced/current/title-21?part=3§ion=3.2
- Rouphael NG, Paine M, Mosley R, Henry S, McAllister DV, Kalluri H, et al. The safety, immunogenicity, and acceptability of inactivated influenza vaccine delivered by microneedle patch (TIV-MNP 2015): a randomised, partly blinded, placebo-controlled, phase 1 trial. Lancet. 2017;390(10095):649-658. doi:10.1016/S0140-6736(17)30575-5. PMID: 28666680.
- Adigweme I, Yisa M, Ooko M, Akpalu E, Bruce A, Donkor S, et al. A measles and rubella vaccine microneedle patch in The Gambia: a phase 1/2, double-blind, double-dummy, randomised, active-controlled, age de-escalation trial. Lancet. 2024;403(10439):1879-1892. doi:10.1016/S0140-6736(24)00532-4. PMID: 38697170.
- Iwata H, Kakita K, Imafuku K, Takashima S, Haga N, Yamaguchi Y, et al. Safety and dose-sparing effect of Japanese encephalitis vaccine administered by microneedle patch in uninfected, healthy adults (MNA-J): a randomised, partly blinded, active-controlled, phase 1 trial. Lancet Microbe. 2022;3(2):e96-e104. doi:10.1016/S2666-5247(21)00269-X. PMID: 35544051.
- van Smeden J, Bouwstra JA. Stratum Corneum Lipids: Their Role for the Skin Barrier Function in Healthy Subjects and Atopic Dermatitis Patients. Curr Probl Dermatol. 2016;49:8-26. doi:10.1159/000441540. PMID: 26844894.
- Wei Q, He Z, Li Z, Zhou Z, Piao Y, Huang J, et al. A skin-permeable polymer for non-invasive transdermal insulin delivery. Nature. 2025;648(8093):459-467. doi:10.1038/s41586-025-09729-x. PMID: 41261125.
- Frum Y, Khan GM, Sefcik J, Rouse J, Eccleston GM, Meidan VM. Towards a correlation between drug properties and in vitro transdermal flux variability. Int J Pharm. 2007;336(1):140-7. doi:10.1016/j.ijpharm.2006.11.049. PMID: 17182201.
- Cheruvu HS, Liu X, Grice JE, Roberts MS. An updated database of human maximum skin fluxes and epidermal permeability coefficients for drugs, xenobiotics, and other solutes applied as aqueous solutions. Data Brief. 2022;42:108242. doi:10.1016/j.dib.2022.108242. PMID: 35599823.
- Choi HK, Acharya G, Lee Y, Lee CH. A Data-Mining Approach for the Quantitative Assessment of Physicochemical Properties of Molecular Compounds in the Skin Flux. AAPS PharmSciTech. 2021;22(3):117. doi:10.1208/s12249-021-01988-x. PMID: 33768360.
- Prausnitz MR, Langer R. Transdermal drug delivery. Nat Biotechnol. 2008;26(11):1261-8. doi:10.1038/nbt.1504. PMID: 18997767.
- He Z, Chen S, Zhang Y, Cheng B, Wan Q, Wang J, et al. ZIF-8-Based Smart-Responsive Nanocomposites for Cutaneous Wound Healing: A Review. ACS Omega. 2026;11(39):58387-58412. doi:10.1021/acsomega.6c05811. PMID: 42851847.
- Yang XX, Feng P, Cao J, Liu W, Tang Y. Composition-Engineered Metal-Organic Framework-Based Microneedles for Glucose-Mediated Transdermal Insulin Delivery. ACS Appl Mater Interfaces. 2020;12(12):13613-13621. doi:10.1021/acsami.9b20774. PMID: 32138507.
- Pastore MN, Kalia YN, Horstmann M, Roberts MS. Transdermal patches: history, development and pharmacology. Br J Pharmacol. 2015;172(9):2179-209. doi:10.1111/bph.13059. PMID: 25560046.
- Hopf NB, Champmartin C, Schenk L, Berthet A, Chedik L, Du Plessis JL, et al. Reflections on the OECD guidelines for in vitro skin absorption studies. Regul Toxicol Pharmacol. 2020;117:104752. doi:10.1016/j.yrtph.2020.104752. PMID: 32791089.
- Parra A, Clares B, Rosselló A, Garduño-Ramírez ML, Abrego G, García ML, et al. Ex vivo permeation of carprofen from nanoparticles: A comprehensive study through human, porcine and bovine skin as anti-inflammatory agent. Int J Pharm. 2016;501(1-2):10-7. doi:10.1016/j.ijpharm.2016.01.056. PMID: 26826569.
- Schreiber S, Mahmoud A, Vuia A, Rübbelke MK, Schmidt E, Schaller M, et al. Reconstructed epidermis versus human and animal skin in skin absorption studies. Toxicol In Vitro. 2005;19(6):813-22. doi:10.1016/j.tiv.2005.04.004. PMID: 15913948.
- Brackin T, Gayes H, Johnson I, Fox D, Roper C. Skin barrier function for regulatory skin absorption tests and effects on testosterone and sucrose absorption. Toxicol In Vitro. 2024;95:105735. doi:10.1016/j.tiv.2023.105735. PMID: 37977296.
- Chakrabarty US, Chakraborty A, Mal M, Sinha D. In Vitro-In Vivo Correlation (IVIVC) in Transdermal Drug Delivery Systems: Models, Regulatory Perspectives and Translational Challenges. Pharm Res. 2026;43(8):2589-2606. doi:10.1007/s11095-026-04181-6. PMID: 42581259.
- Wiśniewska P, Haponiuk J, Saeb MR, Rabiee N, Bencherif SA. Mitigating Metal-Organic Framework (MOF) Toxicity for Biomedical Applications. Chem Eng J. 2023;471. doi:10.1016/j.cej.2023.144400. PMID: 39280062.
- Khan ST, Moosavi SM. Connecting metal-organic framework synthesis to applications using multimodal machine learning. Nat Commun. 2025;16(1):5642. doi:10.1038/s41467-025-60796-0. PMID: 40593657.
- Zheng HQ, Feng HX, Li BX, Hui YF, Lin YH, Su XF, et al. A Zr-based metal-organic framework drug release system with long-lasting antibacterial behavior for accelerating wound healing. Dalton Trans. 2024;53(48):19226-19234. doi:10.1039/d4dt02734e. PMID: 39513398.
- Chen Y, Li P, Modica JA, Drout RJ, Farha OK. Acid-Resistant Mesoporous Metal-Organic Framework toward Oral Insulin Delivery: Protein Encapsulation, Protection, and Release. J Am Chem Soc. 2018;140(17):5678-5681. doi:10.1021/jacs.8b02089. PMID: 29641892.
- Wu T, Hou X, Li J, Ruan H, Pei L, Guo T, et al. Microneedle-Mediated Biomimetic Cyclodextrin Metal Organic Frameworks for Active Targeting and Treatment of Hypertrophic Scars. ACS Nano. 2021;15(12):20087-20104. doi:10.1021/acsnano.1c07829. PMID: 34792332.
- Li S, Long M, Li J, Zhang Y, Feng N, Zhang Z. Improved topical delivery of curcumin by hydrogels formed by composite carriers integrated with cyclodextrin metal-organic frameworks and cyclodextrin nanosponges. Int J Pharm X. 2024;8:100310. doi:10.1016/j.ijpx.2024.100310. PMID: 39717567.
- Peng W, Li J, Cheng J, Yang J, Yang L, Dong H, et al. Core-Shell Microneedle Patch with Controlled Sequential Release of CuO(2)@ZIF-8 Nanoparticles and VEGF for Synergistic Treatment of Infected Wounds. ACS Appl Mater Interfaces. 2025;17(28):40196-40207. doi:10.1021/acsami.5c07830. PMID: 40605239.
- Ettlinger R, Lächelt U, Gref R, Horcajada P, Lammers T, Serre C, et al. Toxicity of metal-organic framework nanoparticles: from essential analyses to potential applications. Chem Soc Rev. 2022;51(2):464-484. doi:10.1039/d1cs00918d. PMID: 34985082.
- Taddei M, Dau PV, Cohen SM, Ranocchiari M, van Bokhoven JA, Costantino F, et al. Efficient microwave assisted synthesis of metal-organic framework UiO-66: optimization and scale up. Dalton Trans. 2015;44(31):14019-26. doi:10.1039/c5dt01838b. PMID: 26165508.
- Yang X, Ruan D, Ye F, Qiu X, Li S, Huang L, et al. Dissolvable silk microneedles delivering minocycline-loaded ZIF-8 nanoparticles for targeted modulation of sebocyte metabolism and improved acne therapy. J Nanobiotechnology. 2026;24(1). doi:10.1186/s12951-026-04221-8. PMID: 41814339.
- Ghaemmaghami P, Aghdam RM, Jamaledin R, Koohikar A, Tafti SHA, Boroumand S, et al. PVA/alginate/maltose microneedle patch loaded with ZIF-8 and Cu-containing ZIF-8 for full-thickness wound healing: In vitro and in vivo studies. Int J Biol Macromol. 2026;369:152402. doi:10.1016/j.ijbiomac.2026.152402. PMID: 42097432.
- Zhang P, Li Y, Tang Y, Shen H, Li J, Yi Z, et al. Copper-Based Metal-Organic Framework as a Controllable Nitric Oxide-Releasing Vehicle for Enhanced Diabetic Wound Healing. ACS Appl Mater Interfaces. 2020;12(16):18319-18331. doi:10.1021/acsami.0c01792. PMID: 32216291.
- Yao S, Wang Y, Chi J, Yu Y, Zhao Y, Luo Y, et al. Porous MOF Microneedle Array Patch with Photothermal Responsive Nitric Oxide Delivery for Wound Healing. Adv Sci (Weinh). 2022;9(3):e2103449. doi:10.1002/advs.202103449. PMID: 34783460.
- Xiao J, Zhu Y, Huddleston S, Li P, Xiao B, Farha OK, et al. Copper Metal-Organic Framework Nanoparticles Stabilized with Folic Acid Improve Wound Healing in Diabetes. ACS Nano. 2018;12(2):1023-1032. doi:10.1021/acsnano.7b01850. PMID: 29406741.
- Sun C, Zheng G, Zhao Y, Liu Z, Jiang Z, Li J, et al. Ultrasound-activated UiO-66 enables ROS-mediated antibacterial therapy and accelerates diabetic wound healing. Regen Biomater. 2026;13:rbag012. doi:10.1093/rb/rbag012. PMID: 41883347.
- Pandya I, Joshi V, Pansuriya R, Raje N, Assiri MA, Malek N. A multifunctional IL@MOF composite-based hydrogel for enhanced transdermal drug delivery of 5-fluorouracil. J Mater Chem B. 2025;13(47):15309-15321. doi:10.1039/d5tb00931f. PMID: 41243804.
- Seifi S, Ghaee A. Fabrication and evaluation of pectin-gelatin skin tissue engineering scaffolds containing amoxicillin trihydrate-loaded UiO-66-NH(2) nanoparticles. Int J Biol Macromol. 2025;320(Pt 3):145904. doi:10.1016/j.ijbiomac.2025.145904. PMID: 40675288.
- Barjasteh M, Vossoughi M, Bagherzadeh M, Pooshang Bagheri K. Green synthesis of PEG-coated MIL-100(Fe) for controlled release of dacarbazine and its anticancer potential against human melanoma cells. Int J Pharm. 2022;618:121647. doi:10.1016/j.ijpharm.2022.121647. PMID: 35288221.
- Márquez AG , Hidalgo T , Lana H , Cunha D , Blanco-Prieto MJ , Álvarez-Lorenzo C , et al. Biocompatible polymer-metal-organic framework composite patches for cutaneous administration of cosmetic molecules. J Mater Chem B. 2016;4(43):7031-7040. doi:10.1039/c6tb01652a. PMID: 32263570.
- Chen G, Leng X, Luo J, You L, Qu C, Dong X, et al. In Vitro Toxicity Study of a Porous Iron(III) Metal‒Organic Framework. Molecules. 2019;24(7). doi:10.3390/molecules24071211. PMID: 30925694.
- Ryu U, Chien PN, Jang S, Trinh XT, Lee HS, Van Anh LT, et al. Zirconium-Based Metal-Organic Framework Capable of Binding Proinflammatory Mediators in Hydrogel Form Promotes Wound Healing Process through a Multiscale Adsorption Mechanism. Adv Healthc Mater. 2024;13(7):e2301679. doi:10.1002/adhm.202301679. PMID: 37931928.
- Zhang XR, Ryu U, Najmiddinov B, Trinh TT, Choi KM, Nam SY, et al. Effect of Silicone Patch Containing Metal-organic Framework on Hypertrophic Scar Suppression. In Vivo. 2024;38(1):235-245. doi:10.21873/invivo.13430. PMID: 38148076.
- Zhai X, Chen K, Wei X, Zhang H, Yang H, Jiao K, et al. Microneedle/CD-MOF-mediated transdural controlled release of methylprednisolone sodium succinate after spinal cord injury. J Control Release. 2023;360:236-248. doi:10.1016/j.jconrel.2023.06.028. PMID: 37355211.
- Coelho LO, Pereira MVS, Andrade TA, Oliveira KM, de Jesus JR. Zinc-Based bioMOF Nanocarrier for Curcumin Release Enables Permeation through Biomimetic Biological Membranes and Controlled Release. ACS Omega. 2026;11(30):44729-44738. doi:10.1021/acsomega.6c00840. PMID: 42569017.
- Xiao J, Chen S, Yi J, Zhang H, Ameer GA. A Cooperative Copper Metal-Organic Framework-Hydrogel System Improves Wound Healing in Diabetes. Adv Funct Mater. 2017;27(1). doi:10.1002/adfm.201604872. PMID: 28729818.
- Bello MG, Zhang J, Chen L. Cyclodextrin metal-organic framework design principles and functionalization for biomedical application. Carbohydr Polym. 2025;364:123684. doi:10.1016/j.carbpol.2025.123684. PMID: 40484611.
- Rizzo M, Marussi G, Crosera M, Marcella M, Biasiol G, Adami G, et al. Gallium and arsenic human skin permeation after application of GaAs particles: an ex-vivo study using Franz cells. Food Chem Toxicol. 2025;206:115758. doi:10.1016/j.fct.2025.115758. PMID: 40992722.
- Liu X, Zhu Y, Jing Y, Zhu Y, Wu L, Nong W. γ-polyglutamic acid-encapsulated metformin-loaded Ga-Car-MOF and hydrocaffeic acid-modified chitosan double-layer microneedle patch for accelerated bacterial-infected wound healing. Int J Biol Macromol. 2025;304(Pt 2):140959. doi:10.1016/j.ijbiomac.2025.140959. PMID: 39952522.
- Wang F, Kong W, Hu W, Zou G, Li Y, Jiang P. A Dual-Layer MOF Microneedle Patch for Self-Sensitized Mild Photothermal Therapy and Autophagy Regulation. Adv Healthc Mater. 2026;15(15):e03743. doi:10.1002/adhm.202503743. PMID: 41635091.
- Chang Z, Wu Y, Hu P, Jiang J, Quan G, Wu C, et al. The Necessity to Investigate In Vivo Fate of Nanoparticle-Loaded Dissolving Microneedles. Pharmaceutics. 2024;16(2). doi:10.3390/pharmaceutics16020286. PMID: 38399340.
- Mitragotri S, Blankschtein D, Langer R. Ultrasound-mediated transdermal protein delivery. Science. 1995;269(5225):850-3. doi:10.1126/science.7638603. PMID: 7638603.
- Mu Y, Fu J, Wang Y, Jin J, You J, Tang D, et al. Oxygen-enriching hollow MOF as nanobubbles to enhance treatment of hypertrophic scar via microneedle-mediated sonodynamic therapy. Biomaterials. 2027;337:124523. doi:10.1016/j.biomaterials.2026.124523. PMID: 42585838.
- Liu L, Zhang H, Peng L, Wang D, Zhang Y, Yan B, et al. A copper-metal organic framework enhances the photothermal and chemodynamic properties of polydopamine for melanoma therapy. Acta Biomater. 2023;158:660-672. doi:10.1016/j.actbio.2023.01.010. PMID: 36640955.
- Liu S, Xin R, Zhang X, Han L. Separable Microneedle Patch Integrated with the Dictamnine-Loaded Copper MOF Nanozyme for Atopic Dermatitis Treatment. ACS Appl Mater Interfaces. 2025;17(18):26386-26401. doi:10.1021/acsami.5c02334. PMID: 40273362.
- Wang H, Fan Z, Zhang J, Zhang L, Gao X, Zhu Y, et al. Bioactive microneedles enabled by Ca-MOF-regulated collagenase for synergistic hypertrophic scar remodeling. Biomater Adv. 2026;185:214897. doi:10.1016/j.bioadv.2026.214897. PMID: 42030680.
- Ge X, Mohapatra J, Silva E, He G, Gong L, Lyu T, et al. Metal-Organic Framework as a New Type of Magnetothermally-Triggered On-Demand Release Carrier. Small. 2024;20(12):e2306940. doi:10.1002/smll.202306940. PMID: 38127968.
- Jayakumar K, Lee YB, Park JP. Stimuli-responsive polymer-metal-organic framework composites for biomedical applications. Adv Colloid Interface Sci. 2026;356:103977. doi:10.1016/j.cis.2026.103977. PMID: 42320080.
- Chai G, Wang N, Xu M, Ma L, Liu X, Ding Q, et al. Poly (vinyl alcohol)/sodium alginate/carboxymethyl chitosan multifunctional hydrogel loading HKUST-1 nanoenzymes for diabetic wound healing. Int J Biol Macromol. 2024;268(Pt 2):131670. doi:10.1016/j.ijbiomac.2024.131670. PMID: 38643919.
- Wang C, Sudlow G, Wang Z, Cao S, Jiang Q, Neiner A, et al. Metal-Organic Framework Encapsulation Preserves the Bioactivity of Protein Therapeutics. Adv Healthc Mater. 2018;7(22):e1800950. doi:10.1002/adhm.201800950. PMID: 30369102.
- Tadepalli S, Yim J, Cao S, Wang Z, Naik RR, Singamaneni S. Metal-Organic Framework Encapsulation for the Preservation and Photothermal Enhancement of Enzyme Activity. Small. 2018;14(7). doi:10.1002/smll.201702382. PMID: 29323458.
- Ji X, Zhou J, Zhou Z, Liu Z, Yan L, Li Y, et al. Recovering skin-nerve interaction by nanoscale metal-organic framework for diabetic ulcers healing. Bioact Mater. 2024;42:112-123. doi:10.1016/j.bioactmat.2024.08.024. PMID: 39280583.
- Li X, Liu J, Chen Y, Zhang J, Li J, Wagner E, et al. Nano/micro-scale targeted RNA delivery systems for cancer therapy. Int J Pharm. 2026;688:126451. doi:10.1016/j.ijpharm.2025.126451. PMID: 41349874.
- Wang F, Miao S, Tang L, Zhang D, Zhang Q, Yin Z, et al. Triphase Interfacial Nucleation Enables Microneedle-Templated Metal-Organic Framework Coating for In Situ DNA Capture and Release. Anal Chem. 2026;98(25):18892-18900. doi:10.1021/acs.analchem.6c01733. PMID: 42315361.
- Nam J, Ahn J, Shin J, Lee N, Lee Y, Cho SM, et al. Transdermal Delivery of an mRNA-Liposome Vaccine via Dissolving Microneedle to Preserve Vaccine Activity and Enhance Immune Activation. Adv Sci (Weinh). 2026;13(34):e22846. doi:10.1002/advs.202522846. PMID: 41955505.
- Xu S, Zhang K, Shao H, Tian Q, Zhang Y, Xu J, et al. An Innovative Sprayable Mg-MOF Hydrogel Enhances Healing of Chronic Wounds with Multifaceted Pathologies in a Porcine Model. ACS Nano. 2026;20(31):21921-21938. doi:10.1021/acsnano.6c05950. PMID: 42579421.
- Sisignano M, Rice ASC, Geisslinger G. Topical Analgesics: Pharmacology and Clinical Applications. Anesthesiology. 2025;143(5):1371-1381. doi:10.1097/ALN.0000000000005579. PMID: 40965053.
- Altharawi A, Aldakhil T, A Alossaimi M. Synthesis of Rh-MOF/PVA-PVP nanofibers for skin cancer and infection inhibition. Front Chem. 2025;13:1575183. doi:10.3389/fchem.2025.1575183. PMID: 40357128.
- Astaneh ME, Hashemzadeh A, Fereydouni N. ZIF-8-based nanomaterials for diabetic wound healing: mechanisms, applications, and future perspectives. J Mater Chem B. 2025;13(46):14931-14948. doi:10.1039/d5tb00194c. PMID: 41150222.
- Zhou X, Feng Z, Liu P, Song Y, Shakir I, Sun X, et al. Mechanisms of MOF-Based Biomaterials in Skin Wound Healing: Antibacterial, Immunomodulatory, Angiogenic, and Neurogenic Effects. Adv Healthc Mater. 2026;15(11):e03944. doi:10.1002/adhm.202503944. PMID: 41420337.
- Sun S, Liu X, Meng X, Yang Z, Zhang X, Dong H. Bimetallic Metal-Organic Framework Microneedle Array for Wound Healing through Targeted Reactive Oxygen Species Generation and Electron Transfer Disruption. ACS Nano. 2025;19(15):15109-15119. doi:10.1021/acsnano.5c02923. PMID: 40208250.
- Abdelkhalek MA, Abdelaal MM, Abdel Gaber SA. Engineered Metal-Organic Frameworks (MOFs)-electrospun nanofibers for wound healing: a review. J Biol Eng. 2026;20(1). doi:10.1186/s13036-026-00650-z. PMID: 41872920.
- Zhu Z, Gong M, Li L, Cui Y. “Self” signal-suppressed metal-organic framework (MOF) nanodrug for enhanced immunotherapy of melanoma via CD47 blockade. Colloids Surf B Biointerfaces. 2026;262:115564. doi:10.1016/j.colsurfb.2026.115564. PMID: 41723989.
- Li C, Zhou S, Chen C, Zhu L, Li S, Song Z, et al. DDTC-Cu(I) based metal-organic framework (MOF) for targeted melanoma therapy by inducing SLC7A11/GPX4-mediated ferroptosis. Colloids Surf B Biointerfaces. 2023;225:113253. doi:10.1016/j.colsurfb.2023.113253. PMID: 36934611.
- Gao W, Song Y, Wu F, Xu S, Liu B, Zeng L, et al. Tumor-Targeted Metal-Organic Framework for Improved Photodynamic Therapy and Inhibited Tumor Metastasis in Melanoma. ACS Appl Mater Interfaces. 2024;16(50):69769-69788. doi:10.1021/acsami.4c18058. PMID: 39652639.
- Chen J, Niu H, Guan L, Yang Z, He Y, Zhao J, et al. Microneedle-Assisted Transdermal Delivery of 2D Bimetallic Metal-Organic Framework Nanosheet-Based Cascade Biocatalysts for Enhanced Catalytic Therapy of Melanoma. Adv Healthc Mater. 2023;12(7):e2202474. doi:10.1002/adhm.202202474. PMID: 36420881.
- Li N, Wan X, Hou X, Zhang J, Zhang C, Sun Y, et al. Cu-Cy Nps@ZIF-8@HA-Mediated Photodynamic Therapy for Cutaneous Squamous Cell Carcinoma. Int J Nanomedicine. 2026;21:560049. doi:10.2147/IJN.S560049. PMID: 41836720.
- Lan S, Zhang J, Li X, Pan L, Li J, Wu X, et al. Low Toxicity of Metal-Organic Framework MOF-74(Co) Nano-Particles In Vitro and In Vivo. Nanomaterials (Basel). 2022;12(19). doi:10.3390/nano12193398. PMID: 36234530.
- Al-Ansari DE, Al-Badr M, Zakaria ZZ, Mohamed NA, Nasrallah GK, Yalcin HC, et al. Evaluation of Metal-Organic Framework MIL-89 nanoparticles toxicity on embryonic zebrafish development. Toxicol Rep. 2022;9:951-960. doi:10.1016/j.toxrep.2022.04.016. PMID: 35875258.
- Jafari S, Izadi Z, Alaei L, Jaymand M, Samadian H, Kashani VO, et al. Human plasma protein corona decreases the toxicity of pillar-layer metal organic framework. Sci Rep. 2020;10(1):14569. doi:10.1038/s41598-020-71170-z. PMID: 32884004.
- Arya J, Henry S, Kalluri H, McAllister DV, Pewin WP, Prausnitz MR. Tolerability, usability and acceptability of dissolving microneedle patch administration in human subjects. Biomaterials. 2017;128:1-7. doi:10.1016/j.biomaterials.2017.02.040. PMID: 28285193.
- Hirobe S, Azukizawa H, Matsuo K, Zhai Y, Quan YS, Kamiyama F, et al. Development and clinical study of a self-dissolving microneedle patch for transcutaneous immunization device. Pharm Res. 2013;30(10):2664-74. doi:10.1007/s11095-013-1092-6. PMID: 23775442.
- Banerjee P, Kollmannsberger KL, Fischer RA, Jinschek JR. Mechanism of Electron-Beam-Induced Structural Degradation in ZIF-8 and its Electron Dose Tolerance. J Phys Chem A. 2024;128(48):10440-10451. doi:10.1021/acs.jpca.4c06391. PMID: 39565713.
- Nguyen HX, Nguyen LT. Economic analysis of microneedle technology: cost-effectiveness, manufacturing scalability, and commercialization challenges. Drug Deliv Transl Res. 2026;16(9):3137-3168. doi:10.1007/s13346-026-02104-x. PMID: 41946826.
- Ahmadi M, Ayyoubzadeh SM, Ghorbani-Bidkorbeh F, Shahhosseini S, Dadashzadeh S, Asadian E, et al. An investigation of affecting factors on MOF characteristics for biomedical applications: A systematic review. Heliyon. 2021;7(4):e06914. doi:10.1016/j.heliyon.2021.e06914. PMID: 33997421.
- Mühlebach S. Regulatory challenges of nanomedicines and their follow-on versions: A generic or similar approach?. Adv Drug Deliv Rev. 2018;131:122-131. doi:10.1016/j.addr.2018.06.024. PMID: 29966685.
- Iftikhar T, Imran T, He J, Zhang C, Xie Y, Ye H, et al. MOFs-on-MOFs modified wearable electrochemical microneedle array for uric acid detection. Biosens Bioelectron. 2026;294:118246. doi:10.1016/j.bios.2025.118246. PMID: 41285075.
- Bhatnagar S, Dave K, Venuganti VVK. Microneedles in the clinic. J Control Release. 2017;260:164-182. doi:10.1016/j.jconrel.2017.05.029. PMID: 28549948.
- Ingrole RSJ, Azizoglu E, Dul M, Birchall JC, Gill HS, Prausnitz MR. Trends of microneedle technology in the scientific literature, patents, clinical trials and internet activity. Biomaterials. 2021;267:120491. doi:10.1016/j.biomaterials.2020.120491. PMID: 33217629.
- US Food and Drug Administration. Considering whether an FDA-regulated product involves the application of nanotechnology: guidance for industry. Silver Spring (MD): FDA; 2014 Jun. https://www.fda.gov/regulatory-information/search-fda-guidance-documents/considering-whether-fda-regulated-product-involves-application-nanotechnology
- European Medicines Agency. Medical devices: EMA’s regulatory role, including medicines used in combination with a medical device and devices made of substances that are systemically absorbed. Accessed 2026-10-30. https://www.ema.europa.eu/en/human-regulatory-overview/medical-devices
- Pétuya R, Durdy S, Antypov D, Gaultois MW, Berry NG, Darling GR, et al. Machine-Learning Prediction of Metal-Organic Framework Guest Accessibility from Linker and Metal Chemistry. Angew Chem Int Ed Engl. 2022;61(9):e202114573. doi:10.1002/anie.202114573. PMID: 34878706.
- Park S, Kim B, Choi S, Boyd PG, Smit B, Kim J. Text Mining Metal-Organic Framework Papers. J Chem Inf Model. 2018;58(2):244-251. doi:10.1021/acs.jcim.7b00608. PMID: 29227671.
- Park H, Kang Y, Choe W, Kim J. Mining Insights on Metal-Organic Framework Synthesis from Scientific Literature Texts. J Chem Inf Model. 2022;62(5):1190-1198. doi:10.1021/acs.jcim.1c01297. PMID: 35195419.
- Ozcan A, Coudert FX, Rogge SMJ, Heydenrych G, Fan D, Sarikas AP, et al. Artificial Intelligence Paradigms for Next-Generation Metal-Organic Framework Research. J Am Chem Soc. 2025;147(27):23367-23380. doi:10.1021/jacs.5c08214. PMID: 40551706.
- Wang H, Yang L, Leng D, Du Y, Ning H. Accelerating the discovery and optimization of metal-organic framework materials via machine learning. Adv Colloid Interface Sci. 2025;346:103671. doi:10.1016/j.cis.2025.103671. PMID: 40957153.
- Sabbagh F, Zakrzewska A, Rybak D, Król J, Abdi A, Nakielski P, et al. Transdermal Drug Delivery Systems Powered by Artificial Intelligence. Adv Healthc Mater. 2025;14(32):e03030. doi:10.1002/adhm.202503030. PMID: 41102882.
- Lv X, Xiang C, Zheng Y, Lv XL, Zhou WX. The Trilogy of Skin Regeneration via Metal-Organic Frameworks Nanomedicine: Precision Management of Refractory Wounds, Pathological Scarring, and Hair Follicle Reactivation. Int J Nanomedicine. 2025;20:10433-10468. doi:10.2147/IJN.S548746. PMID: 40918948.
- Europe PMC. Forward-citation counts for PMID 34783460 (144) and PMID 32138507 (58), retrieved via the Europe PMC REST search API (query CITES:
_MED) on 2026-10-30. https://www.ebi.ac.uk/europepmc/webservices/rest/search