Targeted drug delivery

Targeted drug delivery is the attempt to concentrate a drug in diseased tissue and keep it out of healthy tissue, either by packaging it in a carrier with favourable biodistribution or by attaching it to a molecule that binds a marker on the target cell. It is the oldest ambition in pharmacology and the most persistently disappointing. The therapies that have actually reached patients work by chemistry and antibody biology rather than by the nanoparticle targeting the field spent four decades pursuing.

§How it works

Two mechanisms are available, and they are usually confused.

Passive targeting relies on where a particle goes by default. Carriers between roughly 10 and 200 nanometres circulate longer than free drug, avoid renal filtration, and accumulate preferentially in tissues with leaky vasculature. Tumours were long thought to be such tissues: the enhanced permeability and retention effect, described by Matsumura and Maeda in 1986, holds that fenestrated tumour vessels let macromolecules out and poor lymphatic drainage keeps them there.¹ EPR became the organising rationale for cancer nanomedicine.

Active targeting attaches a ligand — an antibody, an aptamer, a peptide, a sugar — that binds a receptor enriched on the target cell. The intuitive picture is that the ligand steers the carrier. It does not. Studies comparing antibody-targeted and untargeted liposomes found that targeting made no difference to how much drug reached the tumour; what it changed was how much of the drug that arrived got inside cells rather than sitting in the interstitium.² Targeting improves uptake at the destination; it does not improve navigation.

Clearance dominates both. Any particle entering blood adsorbs plasma proteins within seconds, forming a corona that determines its biological identity more than its engineered surface does.³ Opsonised particles are removed by macrophages of the liver and spleen. Polyethylene glycol coatings slow this, and pre-existing anti-PEG antibodies, common in the general population, can accelerate it again.

§The reckoning over EPR

In 2016 a survey of a decade of published animal studies found that the median proportion of an injected nanoparticle dose reaching a solid tumour was about 0.7 percent.⁴ The number was contested — critics argued that percentage of injected dose is the wrong metric, since what matters is concentration at the target relative to toxicity elsewhere — but the central point survived the argument: the accumulation the field had assumed was an order of magnitude smaller than advertised.

A 2020 study then challenged the mechanism itself. Using models in which endothelial gaps could be counted, it concluded that the great majority of nanoparticles entering tumours did so by active transport through endothelial cells rather than by leaking through gaps between them.⁵ If entry is an energy-dependent cellular process rather than passive diffusion through holes, then the design rules derived from EPR — optimise size and circulation time, wait for leakage — were aimed at the wrong variable.

EPR is also far more variable in humans than in the mouse xenografts where it was characterised. Rodent tumours grow fast, are highly vascularised, and are large relative to the animal; human tumours are heterogeneous, often fibrotic, and frequently have high interstitial pressure that opposes convective entry.

Contested

Whether EPR is a usable clinical phenomenon remains disputed. Some groups argue it is real but patient-specific and could be exploited by selecting patients whose tumours show high permeability, using the imaging and circulating-marker methods described in Nanoscale diagnostics. Others argue that the effect is too weak and too inconsistent in humans to build a therapeutic strategy on. No approved nanomedicine's benefit is clearly attributable to EPR.

§What actually worked

Antibody–drug conjugates are the modality that delivered. An antibody against a tumour antigen carries a cytotoxic payload too potent to be given systemically, joined by a linker designed to hold in blood and release inside the cell. Gemtuzumab ozogamicin was approved in 2000, withdrawn in 2010 over toxicity and lack of demonstrated benefit, and reapproved in 2017 at a lower, fractionated dose — a history that illustrates how narrow the therapeutic window is. Brentuximab vedotin, ado-trastuzumab emtansine, and trastuzumab deruxtecan followed, the last extending benefit to tumours expressing only low levels of its target because the released payload diffuses into neighbouring cells.

Conjugation to a sugar ligand achieved something similar for RNA drugs. Attaching a triantennary N-acetylgalactosamine group to a small interfering RNA directs it to the asialoglycoprotein receptor, which hepatocytes display in enormous numbers. The result is subcutaneous dosing, liver-specific silencing, and durable effect from infrequent injections — the most complete example of receptor-directed delivery in medicine.

Nanoparticle carriers succeeded where the goal was reformulation rather than targeting. Liposomal doxorubicin reduced cardiotoxicity. Albumin-bound paclitaxel eliminated a toxic solvent. Lipid nanoparticles made mRNA vaccines and the first systemic in vivo genome-editing therapies possible, and they work because they go to the liver reliably, not because they go anywhere selectively.

What none of these resembles is a carrier that navigates. The addressable containers of DNA nanotechnology come closest to conditional release on a molecular signal, and they have reached mice rather than patients. Where cells must be modified precisely, the field has generally moved the targeting problem out of the body altogether: Casgevy and other ex vivo therapies remove the cells, edit them in a facility, and put them back, which is expensive and immunologically demanding but sidesteps delivery entirely.

§Limitations

The liver is the recurring problem. Its fenestrated endothelium and resident macrophages take up most of what circulates, which is convenient for hepatic targets and an obstacle for everything else. Reformulating a particle to shift its protein corona can redirect it toward lung or spleen, and antibody-decorated particles have reached specific cell types in animals, but no carrier has been approved on the strength of selective delivery to a tissue outside the liver.

The brain is harder still. The blood–brain barrier excludes almost all particles; the approaches with the best evidence use receptor-mediated transcytosis through the transferrin receptor, or temporary mechanical opening of the barrier with focused ultrasound and injected microbubbles, the same modality used for the non-thermal brain stimulation covered in Non-invasive neuromodulation. Both are in clinical development rather than routine use.

Even at the target, a carrier must escape the endosome. For nucleic acid cargoes this is the dominant inefficiency, with only a small percentage of internalised material reaching the cytosol.

Toxicity does not disappear when a drug is targeted. Conjugates cause on-target damage wherever the antigen is expressed, and payload released prematurely in circulation causes conventional cytotoxicity; interstitial lung disease and ocular toxicity have limited specific ADCs. Complement activation by particle surfaces can cause acute infusion reactions unrelated to the drug.

Cost is a structural limitation rather than a technical one. Conjugates and engineered carriers are biologics with complex manufacturing, priced accordingly, and the diffusion problem they create is the subject of Access and inequality. A delivery technology that only works inside well-funded health systems solves a narrower problem than its developers describe.

§Outlook

The lesson the field draws from its own record is that delivery is not a finishing step applied to a finished drug — it determines what drugs are possible. Genome editing is the clearest case: the set of diseases addressable by CRISPR–Cas9 and Base editing as of 2026 is essentially the set of diseases whose causal gene is expressed in hepatocytes, because that is where the carriers go. Extending editing to muscle, immune cells, or the central nervous system is a delivery problem, not an editing one, and the same is true of proposals such as Gene therapy for aging or the systemic administration of Senolytics.

Two directions look most likely to change that. Engineered capsids of the kind described in AAV vectors, and cell-type-selective lipids, are converging on the ability to specify a destination tissue at formulation time, an incremental route that has already produced organ-selective particles in animals and that would widen the reach of Somatic gene therapy more than any change to the editing enzymes themselves. Cell-based carriers — red cell membranes, macrophages, engineered bacteria — borrow biology's own trafficking rather than trying to replicate it, and shade into the externally guided devices covered in Medical microrobots. Neither approach resembles the autonomous targeting machine imagined in Medical nanorobots, and the history of this field is a long argument that the imagined version underrates how thoroughly the body sorts what enters it.

§See also

• Lipid nanoparticles
• Medical nanorobots
• DNA nanotechnology
• Nanoscale diagnostics
• AAV vectors
• Somatic gene therapy
• Medical microrobots
• Access and inequality

§References