From diagnostic PET uptake to predicted therapeutic radiation dose — matched PSMA/DOTATATE imaging-therapy pairs (Ga-68/Cu-64 ↔ Lu-177/Ac-225)
Theranostics begins with a diagnostic positron emission tomography (PET) scan using a short-lived radiometal (gallium-68, half-life 68 minutes, or copper-64, half-life 12.7 hours) chelated to a small targeting molecule — PSMA-11 for prostate cancer or DOTATATE for neuroendocrine tumors. Tracer accumulates in tumor cells proportional to the density of the target receptor, and the resulting standardized uptake value (SUV) becomes a quantitative, non-invasive proxy for how much therapeutic payload the same lesion could later receive.
Standardized Uptake Value normalizes the measured radioactivity concentration in a region of interest to the injected dose and patient body weight:
SUV = (activity concentration in tissue [Bq/mL]) / (injected activity [Bq] / body weight [g])
SUVmax takes the single hottest voxel within a segmented tumor volume — deliberately avoiding partial-volume averaging that would blur small or heterogeneous lesions. For PSMA and SSTR2 (somatostatin receptor type 2) imaging, uptake is receptor-mediated: the tracer binds a specific cell-surface protein overexpressed on tumor cells (PSMA on prostate cancer epithelium; SSTR2 on well-differentiated neuroendocrine tumor cells), so SUVmax is fundamentally a surrogate for receptor density and blood-flow-delivered tracer availability, not tumor size.
Background reference organs matter clinically: liver parenchyma (moderate PSMA/SSTR2 expression) and blood pool are used as internal comparators. A lesion is typically called "positive" when its uptake visibly exceeds liver background — the same qualitative anchor used later for eligibility screening.
PSMA pair (prostate cancer): • Diagnostic: Ga-68-PSMA-11, F-18-DCFPyL, or Ga-68/F-18-PSMA-1007 PET/CT — visualizes prostate-specific membrane antigen, a transmembrane glutamate carboxypeptidase highly overexpressed in prostate cancer (100–1,000× normal prostate tissue) • Therapeutic: Lu-177-PSMA-617 (brand name Pluvicto) — same PSMA-617 targeting ligand, radiolabeled with the beta-emitter lutetium-177
DOTATATE/SSTR pair (neuroendocrine tumors): • Diagnostic: Ga-68-DOTATATE (NETSPOT) or Cu-64-DOTATATE (Detectnet) PET/CT — binds somatostatin receptor subtype 2, overexpressed on well-differentiated NETs (gastroenteropancreatic, bronchial) • Therapeutic: Lu-177-DOTATATE (Lutathera), first PRRT agent FDA-approved (January 2018) based on the NETTER-1 phase 3 trial
In both families, the diagnostic scan is acquired first — always — as a gatekeeper: no PET, no therapy. The entire theranostic paradigm depends on the assumption validated at length in Stage 2: that the diagnostic and therapeutic molecules behave identically in the body.
PSMA and SSTR2 PET are unusual among oncology imaging biomarkers because they are simultaneously diagnostic (staging, restaging, recurrence detection) and dosimetric (predicting therapeutic radiation delivery) — the same scan answers "where is the disease" and "how much therapy will it receive."
The word "theranostic" fuses therapy and diagnostics — and the fusion is chemical, not just conceptual. The diagnostic and therapeutic agents in a theranostic pair are built from the identical targeting peptide or small molecule and, in most pairs, an identical or closely related chelator backbone. Swapping only the radiometal — gallium-68 or copper-64 for imaging, lutetium-177 or actinium-225 for therapy — changes the physics (photon-emitting positron decay versus tissue-damaging beta or alpha particles) without changing where the molecule goes in the body.
A targeting ligand's biodistribution — where it accumulates, how fast it clears, how long it dwells in tumor versus healthy tissue — is governed almost entirely by its molecular structure: the binding affinity of its pharmacophore for the target receptor, its size, charge, and lipophilicity. The radiometal sitting in the DOTA (or DOTAGA) chelator cage contributes negligible mass and, critically, does not participate in receptor binding.
This means Ga-68-PSMA-617 and Lu-177-PSMA-617 are, to a very good first approximation, pharmacokinetic twins: they are taken up by the same PSMA-expressing cells, at the same relative rate, and cleared through the same routes (predominantly renal). The diagnostic scan is therefore not merely "similar" to the future therapy distribution — it is a direct physical preview of it, run days to weeks earlier with a safe, low-radiation-burden isotope.
Small discrepancies do exist: different chelators (DOTA vs DOTAGA vs HBED-CC used in PSMA-11) subtly alter lipophilicity and renal clearance kinetics, and the physical half-life mismatch (Ga-68: 68 minutes vs Lu-177: 6.65 days) means the diagnostic scan captures only an early snapshot while therapeutic dose accumulates over the tracer's much longer residence time. These discrepancies are exactly why Stage 3's statistical regression — rather than a naive 1:1 assumption — is necessary.
Lu-177 decays by beta-minus emission (maximum energy 497 keV, mean tissue penetration range approximately 0.5–2 mm) — a relatively low-energy, longer-range particle well suited to medium-to-large tumor deposits, with concurrent low-energy gamma emissions (113 and 208 keV) that conveniently allow post-therapy SPECT imaging for direct dosimetry verification.
Actinium-225 decays by a cascade of alpha emissions (5.8–8.4 MeV per alpha particle, tissue range only 50–100 micrometers — a few cell diameters). Alpha particles deposit far more energy per unit track length (linear energy transfer, LET) than beta particles, producing far more lethal double-strand DNA breaks per hit. Ac-225-PSMA-617 (targeted alpha therapy, TAT) is under active investigation, particularly for patients whose disease progresses despite Lu-177 therapy, though supply of clinical-grade Ac-225 remains a major bottleneck (produced from Th-229 generators or high-energy proton spallation of thorium/radium targets — global annual production measured in curies, not the kilocuries scale of medical Lu-177).
Because Ac-225's alpha range is only a few cell diameters, it can sterilize micrometastases essentially invisible on PET while sparing more distant normal tissue — but the same short range means heterogeneous receptor expression within a tumor can leave under-dosed regions that a longer-range beta emitter like Lu-177 would still reach by crossfire.
To convert a diagnostic PET number into a therapeutic radiation dose prediction, clinical researchers need paired data: for each of many patients, both the pre-therapy diagnostic SUVmax and the actual tumor absorbed dose delivered during a subsequent therapy cycle, the latter measured via serial quantitative SPECT/CT imaging (typically at 4, 24, 48, and 168 hours post-infusion) fed into MIRD- or voxel-based dosimetry software. Regressing dose against SUVmax across this cohort produces a predictive line — and, like any regression, its reliability (R²) improves as the training cohort grows.
Published studies correlating baseline PSMA PET SUVmax with subsequent Lu-177-PSMA tumor absorbed dose (Kabasakal et al. 2015; Okamoto et al. 2017; Violet trial dosimetry substudies; Hohberg et al. 2016, among others) report a moderate-to-strong positive correlation, but with meaningful scatter — reported correlation coefficients cluster in the range r ≈ 0.5 to 0.85 depending on cohort size, lesion segmentation method, and which SUV metric is used (SUVmax vs SUVmean vs SUVpeak vs total lesion PSMA).
Sources of scatter include: (1) the diagnostic scan is a single early time-point snapshot while therapeutic dose integrates the full time-activity curve (area under the curve) over days; (2) tumor perfusion and receptor internalization kinetics can differ between the short-lived diagnostic tracer and the long-residence therapeutic agent; (3) partial-volume effects disproportionately affect small lesions on PET; (4) inter-patient variation in renal clearance alters whole-body retention and therefore tumor residence time independent of receptor density.
Because of this imperfect correlation, SUV-based dose prediction is used clinically as a triage and hypothesis-generating tool — flagging patients and lesions likely to receive a therapeutic dose — while individualized post-therapy dosimetry remains the gold standard for actually measuring delivered dose.
Regression precision follows the general statistical rule that estimator variance shrinks roughly with the inverse of sample size. A dose-prediction model fit on only 10–15 patient-lesion pairs (typical of early single-center pilot dosimetry studies) will show a wide confidence band and an R² easily swayed by one or two outlier lesions — often an under-perfused necrotic-core metastasis or a lesion with unusually fast washout.
As multi-center registries and larger prospective dosimetry sub-studies accumulate more paired observations — nesting patients across different PSMA ligands, scanners, and reconstruction protocols — the regression line stabilizes and R² rises, though it plateaus well short of 1.0 because some irreducible biological and measurement variance persists no matter how large the cohort. This is precisely the behavior modeled by the "paired datapoints" control in this simulator: R² increases with diminishing returns as n grows from 10 toward 200.
Once a regression relating diagnostic SUVmax to therapeutic absorbed dose has been established across a reference cohort, it can be applied prospectively: take one new patient's own tumor SUVmax from their diagnostic PET, locate that value on the x-axis of the fitted model, and read off the corresponding predicted tumor absorbed dose on the y-axis — before they have received a single millicurie of therapeutic activity.
A predicted dose derived purely from diagnostic SUVmax is a pre-treatment estimate, useful for setting expectations, prioritizing lesions likely to respond, and identifying patients whose overall disease burden looks unlikely to receive a meaningful therapeutic dose despite qualifying on PET positivity alone (for example, very small lesions dominated by partial-volume underestimation).
It is explicitly not a substitute for the definitive post-therapy dosimetry described in Stage 3 — the projected value carries the full uncertainty band of the underlying regression, widened further by any difference between this individual patient's physiology (renal function, tumor perfusion, receptor internalization rate) and the population the regression was trained on. In practice, predicted dose is used to counsel patients, to plan monitoring intensity, and increasingly to inform personalized (rather than fixed) activity prescriptions — administering more or less therapeutic activity per cycle based on the individual's predicted or measured dosimetry rather than the one-size-fits-all 7.4 GBq used in the pivotal VISION trial.
Fixed-activity dosing (same administered radioactivity for every patient, every cycle) is simple and was the regulatory pathway used to approve Lu-177-PSMA-617, but it is a compromise: some patients' tumors are systematically under-dosed relative to their potential benefit, while some patients accumulate unnecessary dose to the kidneys and salivary glands — the two normal organs that most often limit how much activity can safely be given.
Dosimetry-guided personalized prescription flips this: use the patient's own predicted (or, ideally, directly measured via early-cycle dosimetry) tumor and organ doses to titrate subsequent cycle activity — escalating for patients whose tumors are under-dosed relative to normal-organ constraints, and de-escalating or spacing cycles further apart for patients approaching cumulative kidney dose limits (roughly 23 Gy cumulative is a commonly cited soft ceiling, extrapolated from external beam radiotherapy renal tolerance data). Several prospective trials (e.g. personalized dosimetry arms following on from VISION-type fixed-dosing trials) are testing whether this individualization improves the therapeutic index — more tumor kill per unit of kidney/salivary toxicity.
The entire theranostic value proposition rests on this projection step: a five-minute diagnostic PET scan, obtained with no meaningful radiation risk beyond routine imaging, can forecast how a patient's tumor will respond to a therapy that has not yet been given — turning treatment planning from population-average dosing into a patient-specific radiation prescription.
Not every patient — and not every lesion — qualifies for radioligand therapy. Regulatory approval and trial protocols set explicit PET-positivity thresholds: a lesion must show uptake convincingly above a reference background (liver parenchyma for PSMA, or a semiquantitative Krenning score for SSTR imaging) before it is considered a valid target, and the pattern of PSMA-negative but structurally suspicious disease can exclude a patient entirely.
The pivotal VISION trial (Sartor et al., New England Journal of Medicine 2021), which led to FDA approval of Lu-177-PSMA-617 (Pluvicto) for metastatic castration-resistant prostate cancer, used a central, blinded PET read with a specific two-part eligibility rule:
1. PSMA-positive disease: at least one metastatic lesion with uptake exceeding liver parenchyma on Ga-68-PSMA-11 PET/CT 2. No discordant PSMA-negative disease: no lesion that is PSMA-negative (uptake ≤ liver background) but meets defined size/anatomic criteria suggestive of clinically significant disease (e.g. a lymph node >2.5 cm short axis, or a visceral metastasis) — such discordant lesions suggest a tumor clone that would not respond to the PSMA-targeted therapy and would keep growing untreated
This two-part gate is exactly the pattern modeled in this stage's canvas: lesions clearing the uptake threshold pass through as eligible targets, while sub-threshold lesions are flagged as a reason for caution or, in aggregate, exclusion from RLT.
Peptide receptor radionuclide therapy (PRRT) eligibility for neuroendocrine tumors historically used the Krenning score, a simple five-point visual scale developed originally for Octreoscan (In-111-pentetreotide) and carried forward to Ga-68-DOTATATE PET:
• 0 — no uptake above background • 1 — uptake less than liver • 2 — uptake equal to liver • 3 — uptake greater than liver • 4 — uptake greater than spleen or kidney (the two highest normal-organ reference points)
A Krenning score of 3 or 4 in the tumor is generally required for PRRT candidacy — directly paralleling the "uptake exceeds liver background" logic used for PSMA eligibility, and reflecting the same underlying theranostic principle: the diagnostic scan is not just detecting disease, it is functionally testing whether the therapeutic radioligand will find enough receptor density to deliver a meaningful dose.
Roughly 70–90% of patients screened for Lu-177-PSMA-617 in real-world and trial cohorts meet PSMA-positivity criteria, but a clinically important minority are excluded specifically because of PSMA-negative, PET/CT-discordant disease — a direct clinical consequence of the same imaging-therapy matching principle that makes theranostics work in the first place.