When lutetium-177 PSMA-617 and Lu-177 DOTATATE entered routine clinical use, the prescribing convention carried over from decades of empiric radionuclide therapy: a fixed administered activity per cycle, typically 7.4 GBq, with adjustments only for severe renal impairment or prior myelotoxicity. That convention made pragmatic sense when dosimetry was a manual, multi-day calculation that most departments lacked the medical physics capacity to perform. It makes less sense now, and this post explains where the variance lives and how serial SPECT-based PK modeling addresses it.
Why the Same GBq Prescription Delivers Different Doses
Absorbed dose to any organ depends on three independently variable quantities: the amount of radioactivity that localizes there, how long it stays, and the geometry of the source relative to the target tissue. In Lu-177 therapy, all three vary substantially between patients.
Consider two mCRPC patients both receiving 7.4 GBq of Lu-177 PSMA-617. Patient A has high PSMA expression across all metastatic lesions, slow renal clearance, and a lean body habitus. Patient B has moderate expression, rapid urinary excretion, and a heavier build. The same nominal activity delivers tumor-absorbed doses that can differ by a factor of two to three, and kidney doses that diverge comparably. The population-average outcome reported in a phase III trial masks that variance by design: the trial is powered to detect a mean effect, not to characterize the distribution of individual doses.
Published dosimetry series have documented absorbed dose to the kidney ranging from under 2 Gy to over 8 Gy per cycle at a fixed 7.4 GBq, with tumor lesion doses spanning an even wider window. That is not measurement uncertainty; it is true biological heterogeneity between individuals.
The Measurement Foundation: Serial Quantitative SPECT/CT
Patient-specific dosimetry requires quantifying how activity evolves over time in each organ of interest. The practical standard for Lu-177 is serial quantitative SPECT/CT, typically at 24, 96, and 168 hours post-administration, sampling three points on the time-activity curve for kidneys, liver, and tumor lesions.
Two requirements are non-negotiable for the SPECT data to support dosimetric analysis. First, attenuation correction must be applied using the co-registered CT component. Uncorrected SPECT underestimates deep-tissue activity by 30 to 50 percent, which propagates directly into absorbed dose error. Second, recovery coefficient calibration must account for partial volume effects, particularly for lesions under 2 cm where the collimator's point spread function blurs signal into surrounding tissue. Both requirements are supported by the EANM dosimetry guidelines and are feasible in any department running SPECT/CT.
From Time-Activity Curves to the MIRD Pipeline
Once activity concentrations are extracted from each volume of interest at each time point, the MIRD formalism provides the pathway to absorbed dose. The time-integrated activity coefficient (TIAC) for each source organ is calculated by fitting a mono- or bi-exponential function to the serial SPECT data and integrating to infinity. That integral, multiplied by the appropriate S-value for the source-target organ pair, gives mean absorbed dose in gray.
S-values for standard organ geometries are tabulated in OLINDA/EXM and successor software. For kidney dosimetry they are well-validated. For tumor lesions, which are irregular in size and location, voxel-based S-kernel convolution or Monte Carlo transport gives more accurate results, though they require segmented SPECT volumes and corresponding CT anatomy. The key point is that the calculation chain from scan to TIAC to absorbed dose is deterministic once the imaging data are in hand. The uncertainty is not in the physics; it is in whether two or three time points adequately characterize each patient's clearance curve shape.
Where PK Modeling Improves on Standard MIRD Curve Fitting
Standard MIRD-based dosimetry treats the time-activity curve as a measured dataset and fits a simple exponential to it. That approach works when clearance is genuinely mono-exponential and all three time points are available without significant acquisition artifacts. In practice, neither is always true.
Renal clearance of Lu-177 radiopharmaceuticals typically shows an early rapid washout phase followed by a slower terminal phase, reflecting redistribution from blood and soft tissue. A two-point measurement sampling only the late phase will overestimate the TIAC unless the early phase is accounted for. A compartmental PK model, parameterized to each patient's scan series, imposes physiologically plausible structure on the clearance trajectory and uses all available data to constrain it simultaneously rather than fitting each time point independently.
In our pilot work across 74 patient scans at three oncology centres, the kidney TIAC estimated from a two-compartment PK model differed from a simple bi-exponential fit to the same data by a median of 12 percent, with individual cases reaching 30 percent divergence. At a cumulative kidney dose constraint of 23 Gy across a treatment course, a 12 percent systematic error in TIAC translates to roughly a 2.8 Gy discrepancy in cumulative dose estimate. That is clinically meaningful when deciding whether a fourth cycle is within tolerance.
Translating Absorbed Dose Back to Clinical Decisions
The practical output of patient-specific dosimetry is not a number for its own sake. It is a cycle-by-cycle kidney dose estimate and a projected cumulative dose toward the 23 Gy constraint. That projection changes the prescribing question from "how many cycles of standard 7.4 GBq can I give" to "at what activity per cycle does this patient stay within renal constraint while maximizing tumor delivery."
For patients with fast renal clearance, this sometimes supports prescribing above the fixed 7.4 GBq, increasing tumor dose without exceeding the renal constraint. For patients with slow clearance or compromised renal reserve, it identifies early that the standard dose is already marginal and that dose adjustment or cycle spacing modification is warranted before toxicity appears rather than after.
It is worth stating clearly: patient-specific dosimetry does not replace clinical judgment. It provides better absorbed dose information as input to that judgment. The oncologist still decides on cycle number, monitoring intervals, and whether a given patient profile warrants protocol modification. What changes is the quality of the dose information available at each decision point.
The Path to Routine Clinical Dosimetry
The barrier to implementing patient-specific dosimetry in most theranostics programs is not the underlying physics or the imaging equipment. It is workflow integration: coordinating three SPECT acquisitions across five to seven days post-administration, extracting organ volumes consistently, and producing a dosimetry report in time to inform the next cycle prescription.
That coordination burden is what the YSOTOPE platform is built to remove. Serial SPECT uploads, organ segmentation, PK model fitting, and dose report generation run in one connected pipeline. A nuclear medicine physician reviewing the output examines a dose summary, not a multi-step calculation workflow. The data collection occurs in parallel with the patient's normal post-administration monitoring visits, adding no additional imaging appointments to the treatment schedule. The infrastructure for routine dosimetry exists. The question for each department is whether the operational pathway is in place to use it.