What biodistribution means
Biodistribution is the spatial and temporal answer to a simple question: after a compound is administered, where is it, and how much of it is there, at each point in time. For most drugs that answer is inferred from plasma concentrations and a handful of tissue samples. For a radiopharmaceutical it can be measured directly in a living subject, because the molecule carries its own signal.
The useful unit is a fraction, not a concentration. Preclinical biodistribution studies report percentage of injected activity per gram of tissue (%ID/g) at set times after administration, which makes tumour and healthy-tissue uptake directly comparable. The shape of those curves over time matters as much as their peaks: a tissue that takes up little but clears slowly can end up with more accumulated exposure than a tissue that takes up a lot and empties.
Biodistribution is therefore pharmacokinetics with a map attached. It is decided by the ligand's size and charge, by its route of clearance, by the expression of its target in healthy tissue, and by non-specific mechanisms that have nothing to do with the target — filtration and tubular reabsorption in the kidney being the most consequential one for peptides.
What a PET or SPECT image actually shows
PET and SPECT do not image the drug. They image decays. In PET, a positron from the tracer annihilates with an electron and the two resulting 511 keV photons are detected in coincidence, which places that decay on a line between two detectors; reconstruction turns millions of those lines into a map of activity concentration. In SPECT, a rotating gamma camera with a lead collimator accepts photons arriving from one direction at a time, and the projections are reconstructed the same way.
What comes out is a three-dimensional array of activity concentration, in becquerels per millilitre. The clinically familiar whole-body view — the rotating dark-on-light image a physician reads first — is a maximum-intensity projection through that array, and PET/CT pairs it with a CT for anatomy and for attenuation correction.
Because the same ligand can be labelled for imaging or for therapy, this measurement is what makes the modality self-reporting. The distribution seen with the diagnostic label is the best available prediction of where the therapeutic label will go — with the caveat that the two radionuclides have different half-lives, so the imaging window and the therapeutic residence time are not the same.
SUV, in plain language
Raw activity concentration is not comparable between patients: a bigger person and a larger injected activity both inflate it. The standardized uptake value normalises both away. It is the decay-corrected activity concentration measured in a region of interest, divided by the injected activity per gram of body weight.
That definition has one property worth memorising. If all of the injected tracer stayed in the body and spread perfectly evenly, the SUV everywhere would be 1. So an SUV is a ratio to a fair share: an SUV of 1 means a tissue holds exactly its share of the tracer, and an SUV of 8 means it holds eight times its share. A lesion that is obviously "hot" on a display is a region whose SUV is several times the surrounding tissue's.
Two practical cautions come with it. SUVmax — the single brightest voxel in a region — is more reproducible than the region's mean, because it does not depend on exactly where the boundary is drawn, which is why it is what most reports quote. And an SUV is only comparable to another SUV if the whole chain that produced it was consistent: scanner calibration, uptake time, reconstruction settings, and how the region was drawn all move the number. Kinahan and Fletcher's review of the sources of bias and variance is the standard reference for that, and its conclusion is that SUVs are useful for following one patient over time provided nothing in the methodology changed.
Why absorbed dose to healthy organs is monitored
Dosimetry is the step that turns the picture into a number a clinician can act on: absorbed dose, the energy deposited per unit mass of tissue, in gray. It is computed from the time-integrated activity in a region and a model of how the radionuclide's emissions deposit energy there.
Where this matters in practice is the organs at risk. The EANM dosimetry committee's recommendations for lutetium-177-labelled somatostatin-receptor and PSMA ligands exist specifically to encourage patient-specific dosimetry, and summarise the evidence for kidneys, bone marrow, salivary glands, lacrimal glands, pituitary glands, tumours, and the skin in the event of extravasation. The kidney, that document notes, "is generally considered the dose-limiting organ" for somatostatin-receptor therapy, because of non-specific uptake by proximal tubular cells.
The honest part of that guidance is how provisional the limits are. The constraints used in dosimetry-guided trials are extrapolated from external-beam radiotherapy, and the 2 Gy bone-marrow threshold for severe haematological toxicity is "generally considered" to hold by analogy with iodine-131 therapy, with the committee stating that "confirmation of this 2 Gy threshold is still needed" for lutetium-177 agents. Thresholds borrowed from a different way of delivering radiation are the field's working assumption, not a settled fact.
Whether dosimetry is performed at all varies by jurisdiction and by intent. The joint EANM/SNMMI procedure guideline for lutetium-177 PSMA therapy records that the agent was approved using fixed standard treatment activities and that "patient individual dosimetry is not mandatory for in-label use", while recommending it where an individualised regimen is the goal. In much of the European Union a post-therapy scan for verification is effectively expected under the Euratom basic safety standards, even when full dosimetry is optional.
For a drug that has not been approved yet, the requirement is firmer. US regulation obliges an application for a radioactive drug to carry "sufficient data from animal or human studies to allow a reasonable calculation of radiation-absorbed dose to the whole body and critical organs", and requires Phase 1 studies of radioactive drugs to generate data adequate for dosimetry. FDA's 2025 draft guidance on dosage optimisation for therapeutic radiopharmaceuticals goes further and asks that every novel molecular entity have dosimetry studies performed early in clinical development.
Why predicting the distribution early de-risks a programme
Everything above is measured after a molecule exists and has been given to something living. That is late. By then the target has been chosen, the linker and chelator have been designed, radiochemistry has been developed, and a budget has been spent — and the answer that arrives is frequently the one that ends the programme.
This is why physiologically based pharmacokinetic modelling, and more recently machine learning on top of it, has become an active research area in nuclear medicine: PBPK models of organ compartments joined by flows have been used to predict organ and tumour uptake of somatostatin-receptor agents, and PBPK-constrained deep learning has been applied to pre-therapy prediction of voxel-level dosimetry. The published limit of that work is worth keeping in view: most of it is validated in silico or against virtual patients, with clinical voxel-level validation still emerging.
Two other phenomena argue for asking early rather than measuring late. The tumour sink effect means the same administered activity exposes healthy organs differently depending on how much tumour there is to absorb it — so a healthy-tissue estimate from one cohort does not transfer cleanly to another. And a large fraction of a candidate's whole-body exposure is decided by properties chosen at design time — molecular size, charge, clearance route — not by anything that can be tuned afterwards.
Epitorix's own position on this is stated with its fence: in retrospective analyses of radiopharmaceuticals with documented clinical outcomes, its models detected previously reported development and safety liabilities with up to 93 % accuracy. The method beat is on the science page; the site's 3D body chapter renders a published physiological FDG SUV distribution, credited on the credits page, beside one candidate's predicted organ-risk profile. A predicted profile is a hypothesis for the imaging study to test, and the site says exactly that.
