Dosimetry for Radiopharmaceutical Therapy

Abstract

Radiopharmaceutical therapy (RPT) involves the use of radionuclides that are either conjugated to tumor-targeting agents (eg, nanoscale constructs, antibodies, peptides, and small molecules) or concentrated in tissue through natural physiological mechanisms that occur predominantly in neoplastic or otherwise targeted cells (eg, Graves disease). The ability to collect pharmacokinetic data by imaging and use this to perform dosimetry calculations for treatment planning distinguishes RPT from other systemic treatment modalities. Treatment planning has not been widely adopted, in part, because early attempts to relate dosimetry to outcome were not successful. This was partially because a dosimetry methodology appropriate to risk evaluation rather than efficacy and toxicity was being applied to RPT. The weakest links in both diagnostic and therapeutic dosimetry are the accuracy of the input and the reliability of the radiobiological models used to convert dosimetric data to the relevant biologic end points. Dosimetry for RPT places a greater demand on both of these weak links. To date, most dosimetric studies have been retrospective, with a focus on tumor dose-response correlations rather than prospective treatment planning. In this regard, transarterial radioembolization also known as intra-arterial radiation therapy, which uses radiolabeled (⁹⁰Y) microspheres of glass or resin to treat lesions in the liver holds much promise for more widespread dosimetric treatment planning. The recent interest in RPT with alpha-particle emitters has highlighted the need to adopt a dosimetry methodology that specifically accounts for the unique aspects of alpha particles. The short range of alpha-particle emitters means that in cases in which the distribution of activity is localized to specific functional components or cell types of an organ, the absorbed dose will be equally localized and dosimetric calculations on the scale of organs or even voxels (~5 mm) are no longer sufficient. This limitation may be overcome by using preclinical models to implement macromodeling to micromodeling. In contrast to chemotherapy, RPT offers the possibility of evaluating radiopharmaceutical distributions, calculating tumor and normal tissue absorbed doses, and devising a treatment plan that is optimal for a specific patient or specific group of patients.

Introduction

Radiopharmaceutical therapy (RPT) involves the use of radionuclides that are either conjugated to tumor-targeting agents (eg, nanoscale constructs, antibodies, peptides, and small molecules) or concentrated in tissue through natural physiological mechanisms that occur predominantly in neoplastic cells. In the latter category, radioiodine therapy of thyroid cancer is the prototypical and most widely implemented RPT. In the category of radionuclide-ligand conjugates, antibody and peptide conjugates have been studied extensively.^1–4 The efficacy of RPT relies on the ability to delivery cytotoxic radiation to tumor cells without causing prohibitive normal tissue toxicity. After some 30 years of preclinical and clinical research, a number of recent developments suggest that RPT is poised to emerge as an important and widely recognized therapeutic modality. These developments include the substantial investment in antibodies by the pharmaceutical industry and the compelling rationale to build upon this already existing and widely tested platform. In addition, the growing recognition that the signaling pathways responsible for tumor cell survival and proliferation are less easily and durably inhibited than originally envisioned has also provided a rationale for identifying agents that are cytotoxic rather than inhibitory.^5–8 The success and recent Food and Drug Administration approval of the alpha-emitter ²²³Ra ([²²³Ra]RaCl₂, a calcium mimetic) for patients with prostate cancer with castration-resistant skeletal metastases⁹ has also been an important impetus for reconsideration of RPT.

The recent interest in RPT with alpha-particle emitters has highlighted the need to adopt a dosimetry methodology that specifically accounts for the unique aspects of alpha particles.¹⁰ Accordingly the review initially focuses on therapeutic dosimetry for low linear energy transfer (LET) emitters, beta particles, and photons. The dosimetry associated with alpha-particle emitters is considered separately. We begin with a discussion of dosimetry for low LET RPT.

Treatment Planning in RPT

The ability to collect pharmacokinetic (PK) data by imaging and use this to perform dosimetry calculations for treatment planning distinguishes RPT from other treatment modalities. The significance of this may be understood by comparison with chemotherapy dosing. Chemotherapy is administered on a per body weight or body surface area basis. Typically the dose used is based on a phase I dose escalation trial that determines the maximum tolerated dose (MTD). The MTD is defined by the response of a limited number of patients, typically 6 patients in a dose group. This MTD is then used to treat all other patients in phase II and subsequent trials. This approach does not account for differences in drug clearance, metabolism, or PK in different patients. The outcome of such an approach is that patients will either be underdosed (in RPT this is typically the case, as the MTD, when specified in terms of administered activity (AA) is conservatively chosen to avoid toxicity across most patients) or be overdosed and experience toxicity. In RPT, it is possible to collect PK and imaging data to calculate tumor and dose-limiting organ (DLO) absorbed dose (AD). AD is most closely related to tissue damage and normal organ toxicity. Thus AA can be adjusted to customize treatment for each patient to deliver the maximum possible AD to tumor without exceeding the AD to the DLO.

Matching the Dosimetry Methodology to the End Point

Such a scheme has not been widely adopted, in part, because early attempts to relate dosimetry to outcome were not successful. Dosimetry for RPT requires a fundamentally different approach to that used for diagnosis. In diagnostic nuclear medicine, dosimetry is used primarily to evaluate the risk of cancer associated with the imaging procedure. In this context, the mean organ AD to a well-defined anatomical model is needed, as risk data are expressed in terms of mean organ AD to a model that is representative of the exposed (or imaged) population. In therapy, the relevant end points are organ toxicity and tumor control for the individual patient. This end point requires dosimetry that is substantially more detailed than the average over an organ or tumor volume. This information is required because the radiobiological models that yield toxicity and tumor control require dose-volume histograms and knowledge of the dose distribution to sensitive or dose-limiting portions of the tissue. These distinctions in input, dosimetry methodology, radiobiological modeling, and end points are summarized in Figure 1. The Table defines the quantities listed in the figure.

Figure 1.

Table1.

Quantity	Description
à (rS)REF	Time-integrated activity in source region, S, of a reference anatomical model
D(rT)REF	Absorbed dose to target region, T, of a reference anatomical model
wR and wT	Radiation and tissue-weighting factors, respectively
H (rT), E, and LAR	Equivalent dose to target region, T, of a reference model, effective dose to a reference model, and lifetime attributable risk, respectively
A(x,y,z,t), ρ(x,y,z,t), and Z(x,y,z,t)	Spatiotemporal map of the activity, tissue density, and tissue composition (atomic number [Z]-value), respectively
(x,y,z,t), D(x,y,z), and DVH(ν)	Absorbed dose-rate, absorbed dose, and dose-volume histogram for a patient-specific tissue volume, v, respectively
α and β	Tissue-specificcoefficients of radiation damage proportional to dose (single event is lethal) and dose squared (2 sublethal events required for lethal damage), respectively
μ and RBE	DNA repair rate assuming exponential repair of DNA damage, relative biological efficacy
BED, NTCP, and TCP	Biological effective dose, normal tissue complication probability, and tumor control probability, respectively

Dosimetry Methodology for Therapy End Points

As shown in Figure 1, a single value, the time-integrated activity (TIA) in each source organ is required as input into an AD calculation for diagnostic imaging wherein the end point is cancer and health detriment risk. The TIA is then used as an input into generic human models with representative anatomy. By contrast, evaluation of therapeutic end points, under most circumstances, uses the full voxelized information available Q11 from single photon emission CT (SPECT) and PET images, and also requires a registered CT scan, which provides the detailed patient-specific anatomical information, including the tumor. SPECT or PET provide the spatial activity information at different times following injection of the therapeutic radiopharmaceutical. The CT scan can be converted to a density map. The density map may be used to generate a tissue composition map by segmenting air, tissue, and bone according to standard density thresholds.¹¹,¹² The density, composition, and activity maps are input to a Monte Carlo calculation (per time point) to obtain the patient-specific output indicated in Figure 1. If the differences in density and tissue composition are negligible, voxelized S-value and pointkernel methods may be used.¹³,¹⁴ Techniques have also been developed to use voxelized S-value and point-kernel methods by making first-order adjustments to the AD map that account for a range of density variations.¹⁵ The dosimetry output is converted to organ toxicity and tumor control using radiobiological models, which are largely based upon experience in external beam radiotherapy and brachytherapy.^16–19

In the course of developing a new RPT, the best and most cost-effective time to implement patient-specific, 3-dimensional (3D) imaging–based methods is in the context of a phase I trial. The clinical end point for phase I trials is toxicity evaluation, other end points, including collection of imaging data for dosimetry and PK analysis, are also typically included to better understand the treatment. A simplified approach to incorporating treatment planning and 3D-based dosimetry methods in RPT is illustrated in Figure 2. In preclinical testing, the DLO and maximum tolerated AD are determined for a relevant model system. The AD to the dose-limiting or critical tissue is more closely related to potential toxicity than the AA. In such an approach, AA is increased in a relevant animal model until toxicity is reached. The organ responsible for dose-limiting toxicity is identified (eg, by necropsy or histopathology or both), and the AD to this organ at the toxic AA is calculated. This information is used to design a phase I trial in which the escalation variable is the AD to the DLO. In such a trial the “dose” of activity administered is determined by a pretreatment imaging study that is used to obtain the PKs needed to calculate the AD to the DLO. At each AD level in the escalation scheme, the AA required to deliver the AD for each patient in the particular AD cohort is calculated. This means that patients in the same dose (AD) cohort may receive different AAs. In addition to providing the maximum tolerated AD, analysis of the data resulting from such a phase I study design will provide valuable information for future trials, especially in combination therapy investigations. In particular, if analysis of these data suggests that the advantage of treatment planning–based RPT is minimal for the given RPT and patient population, then treatment based on AA may be adequate as the AD delivered to the DLO is not highly dependent on individual patient’s anatomy or PKs.

Figure 2.

Limitations and Uncertainties

The weakest links in both diagnostic and therapeutic dosimetry are the accuracy of the input and the reliability of the radiobiological models used to convert dosimetric data to the relevant biologic end points. Dosimetry for RPT places a greater demand on both of these weak links. Typically, the activity distribution input is at the voxel or multi-voxel level. Reliable results at this level require correction for partial volume and spill-in or spill-out effects as well as corrections for scatter and attenuation.^20–24 Experience with the software package, 3D-radiobiological dosimetry, has shown that voxelized AD, averaged over an organ volume compares very well with the average absorbed-dose-to-an-organ volume.^25,26 The former is obtained by calculating the ADto each voxel and then taking the average of all AD values in a collection of voxels that define an organ volume. The latter is obtained directly from a Monte Carlo calculation in which the energy deposited to the organ volume is tallied and then divided by the mass of the volume (as obtained from the density map).

Radiobiological models in radiopharmaceutical dosimetry present a more fundamental problem particularly because their use requires parameter values that are almost always approximations of their true value. The radiation and tissue-weighting factors for cancer risk end points represent consensus values obtained by Committee review of relevant epidemiologic and biological data.^27–29 In therapy, although there is a consensus of appropriate values for different organ and tumor types, there is no standardization of values analogous to that in the radiation protection realm. In part, this is because the radiobiological parameter values relevant to therapy depend on a host of complex factors. In normal tissue, these factors include organ architecture, treatment history of the patient, and the general metabolic and physiological state of the organ. Radiobiological parameter values in tumors are affected by tumor type and by a number of tumor biology-specific factors that include the hypoxic and metabolic or proliferative fraction of the tumor. Depending upon the scale of the dosimetry calculation, the radiobiological parameters will vary across the normal tissue or tumor volume. The radiosensitivity of different normal tissue components is already known to be spatially dependent.^30–32 This is particularly important for RPT wherein the agent may concentrate within a subset of the cells making up a particular organ. Currently, dosimetry calculations of the type shown in Figure 1 for therapy assume a single set of radiobiological parameter values for all cells making up a tumor or normal organ.

From a practical standpoint, implementation of AD-based treatment planning requires the ability to acquire the necessary images from pretherapeutic administrations and perform the dosimetric calculations, including the Monte Carlo simulations, in a clinical time frame.³³ To date, most dosimetric studies have been retrospective, with a focus on tumor dose-response correlations rather than prospective treatment planning, although a number of schemata for such treatment planning have been proposed.^34–40 In this regard, transarterial radioembolization (TARE) also known as intra-arterial radiation therapy, which uses radiolabeled (⁹⁰Y) microspheres of glass or resin to treat lesions in the liver holds much promise for more widespread dosimetric treatment planning.^41–44 As the spheres stick (embolize) in the vasculature of the liver and tumors, there is no relocalization of activity typical of RPT and a single image suffices to provide the requisite dosimetric guidance. The growing interest in TARE dosimetry has also been fueled in large part by the recently discovered ability to image ⁹⁰Y with either (1) PET owing to a low branching ratio (32 ppm) of internal conversion positron production,⁴⁵,⁴⁶ or (2) SPECT imaging of the bremsstrahlung photons.⁴⁷ The caveat for TARE treatment planning, however, is the difference in the nature of the pretherapeutic radiopharmaceutical, ^99mTc-labeled macroalbumin aggregate, which could potentially have a different biodistribution than the therapeutic microspheres. The predictive ability of the pretherapeutic macroalbumin aggregate remains controversial.^48–50

Dosimetry for Alpha-Particle Emitting Radiopharmaceuticals

This topic was recently extensively reviewed.^10,51,52 In this section, we present specific examples that illustrate the dosimetry approach required for alpha-particle emitter RPT.

Alpha-particle emitters are attractive alternatives to traditional RPT or chemotherapy as a systemic cancer treatment because of their high-energy deposition (LET) over a relatively short range (50–100 μm). The number of hits for cell kill is orders of magnitude fewer than for traditional RPT (3–4 for alphas vs thousands for betas). This is due to the higher LET, but also because the biological effect of alpha radiation differs from other, previously used types of radiation such as beta-particle emitters. AD from alpha emitters is approximately 5 times more potent than AD from beta emitters or external beam radiation¹⁰; this is termed the relative biological effectiveness (RBE). Although a value of 5 is nominally used, the RBE can change substantially if modulators of DNA repair pathways are used.⁵³

High LET radiation is also more damaging to normal organ cells; moreover, the short range of alpha-particle emitters means that in cases in which the distribution of activity is localized to specific functional components or cell types of an organ, the AD will be equally localized and dosimetric calculations on the scale of organs or even voxels (~5 mm) are no longer sufficient. Currently, no modalities are available to image activity at that scale clinically; therefore, models are being developed to supplement the information provided by the SPECT or PET images. The principle is to measure microdistribution of activity over time specific to an organ and radiopharmaceutical using preclinical ex vivo data imaged with an alpha-camera.⁵⁴ Based on these measurements, an apportionment of TIA to substructures can be made using traditional macroscopic imaging (macromodeling to micromodeling). TIA is then converted to AD apportioned to the different organ substructures based on Monte Carlo simulations of radioisotope decay using idealized anatomical models of the organ substructures. This technique is highly analogous Q15 to the traditional Medical Internal Radiation Dose absorbed fraction methodology^55–57 as well as the kidney subunit model (with delineation of cortex, medulla and renal pelvis) designed with peptide receptor radiation therapy in mind.^31,32 Several examples of this alpha-particle dosimetry modeling paradigm have already been published,^58–60 2 of which we present here briefly.

The model for the kidney was inspired by preclinical results using ²²⁵Ac-labeled antibody to treat metastatic breast cancer to the lungs.⁶¹ The treatment showed efficacy, but renal toxicity was observed at average ADs (~2 Gy) well below those consistent with toxicity, even taking into account the RBE. A nephron-based model, with compartments for the glomeruli, believed to be the more radiosensitive cells in the kidney, as well as a potential activity localization point, along with the proximal tubule cells and the distal tubule cells. S-values for the compartments and the different potential isotopes were calculated.⁵⁸

A different scenario holds true for clinical results from the [²²³Ra]RaCl₂ therapy of prostate cancer bone metastases. Traditional absorbed fraction dosimetry predicts a much higher level of hematotoxicity than experienced clinically⁹; which, although beneficial for the patients, impedes the possibility of any dosimetric-based treatment planning. Here, a simple trabecular marrow cavity model was proposed, wherein the fraction of marrow volume receiving a cytotoxic AD was calculated. The discrepancy between an absorbed fraction methodology and a method that specifically accounts for the dose distribution within the marrow cavity could then be explained because a substantial portion of the marrow cavity was not being irradiated.⁵⁹

Other approaches for alpha-particle dosimetry, generally microdosimetric in nature, have been proposed and are being developed^62–67; however, the advantage of the small-scale modeling approach is the ability to be applied to clinical imaging and dosimetry.

Future Directions

Several factors favor the adoption of AD-based treatment planning in the years to come: (1) the more widespread use of combined 3D imaging modalities, such as SPECT/CT, PET/CT (and PET/MRI), as well as the improved reconstruction techniques now available that together provide much more accurate and reliable quantitative data as input^68,69; (2) the development of more 3D dosimetry packages, which are able to exploit such data^70–73; and (3) the focus by the National Institutes of Health on the development of more personalized medicine.⁷⁴

Additionally, the encouraging clinical and preclinical results from alpha-particle emitter RPT foreshadow a greater role to be played by alpha emitters in the fight against cancer, although the alpha-dosimetry models are still in their early stages and application of dosimetric treatment planning for alpha-particle emitter RPT is an important research area. In this context, the stochastic nature of alpha-emitter radiation, due to the short range and high potency at relatively low number of hits has not yet been well integrated into the dosimetry models and need to be better understood.

Finally, in the war on cancer, successful therapies often depend upon combination strategies involving different modalities. Development and use of dosimetric paradigms that will enable optimal integration of the different modalities will play a pivotal role in the success of combined RPT-RPT strategies,^36,75–77 as well as for combinations with other modalities.^35,40,53,78