Visual Abstract
Abstract
In a recent multicenter study, discrepancies between PET/CT-measured activity and vendor-calibrated activity for 90Y glass and resin microspheres were found. In the present work, the origin of these discrepancies was investigated by Monte Carlo (MC) simulations. Methods: Three vial configurations, containing 90Y-chloride, 90Y-labeled glass microspheres, and 90Y-labeled resin microspheres, were modeled with GAMOS, and the electric signal generated in an activity meter was simulated. Energy deposition was scored in the activity meter–active regions and converted into electric current per unit activity. Internal bremsstrahlung (IB) photons, always accompanying β-decay, were simulated in addition to 90Y decays. The electric current per source activity obtained for 90Y glass and resin microspheres, Iglass and Iresin, was compared in terms of relative percentage difference with that of 90Y-chloride ( and
) and each other (δ). The findings of this work were compared with the ones obtained through PET measurements in the multicenter study. Results: With the inclusion of IB photons as primary particles in MC simulations, the
and
results were 24.6% ± 3.9% and −15.0% ± 2.2%, respectively, whereas δ was 46.5% ± 1.9%, in very good agreement with the values reported in the multicenter study. Conclusion: The MC simulations performed in this study indicate that the discrepancies recently found between PET/CT-measured activity and vendor-calibrated activity for 90Y glass and resin microspheres can be attributed to differences in the geometry of the respective commercial vials and to the metrologic approach adopted for activity meter calibration with a 90Y-chloride liquid source. Furthermore, IB photons were shown to play a relevant role in determining the electric current in the activity meter.
The use of 90Y for radionuclide therapy in nuclear medicine considers a variety of therapeutic options: radiosynoviorthesis (1), anti-CD20 antibodies (2), radiolabeled peptides (3), and selective internal radiotherapy with glass and resin 90Y-labeled microspheres (4).
In particular, 90Y transarterial radioembolization by permanent implantation of 90Y-labeled microspheres is an established therapeutic option for unresectable primary hepatic carcinoma and liver metastases (4–6). Selective internal radiotherapy has been applied for 2 decades, and hundreds of thousands of patients have benefited from this treatment so far, worldwide. Clear evidence of the superiority of dosimetry-based treatment personalization is present in the literature (7); hence, treatment personalization of the administered therapeutic activity, based on predictive dosimetry, is now broadly recommended.
An accurate determination of the therapeutic administered activity is key in dose–response studies aiming at optimization of treatment safety and efficacy and in view of possible combined internal and external irradiator therapeutic scenarios (8–10). Presently, a quantitative accuracy within 10% is considered the minimum standard for assessment of a therapeutic activity administration (11).
Evidence of possible discrepancies from the vendor-declared calibrated activity was found for the resin spheres (12). Recently, Gnesin et al. (13), in a multicenter and multidevice investigation, reported significant discrepancies between PET/CT-measured activity and vendor-calibrated activity for both 90Y microsphere devices. Furthermore, 90Y activities assessed with PET measurements were found to systematically underestimate the vendor-calibrated activity for glass spheres (−21% on average) and overestimate for the resins (+15%). The relative activity difference between resin and glass microspheres averages 46%. Interestingly, good agreement between PET- and vendor-calibrated activities was found for a set of vials containing a homogeneous 90Y-chloride liquid solution.
A possible explanation for the observed discrepancies could reside in differences in geometric configuration and material composition across the 3 commercial products (90Y-labeled resin and glass microsphere vials and the vial containing 90Y in a homogeneous liquid chloride solution). In this work, we performed Monte Carlo (MC) simulations to model the output signal from a reference activity meter in the 3 commercial 90Y vial configurations. We compared the MC results with the recently published experimental results (12,13).
MATERIALS AND METHODS
We hypothesized that the differences in geometric configuration and material composition across the 3 vial configurations potentially lead to an important difference in the emitted bremsstrahlung energy spectrum contributing to the signal generation in activity meters. Consequently, activity for glass and resin microsphere vials is misestimated adopting the configuration typically used at the manufacturer site before shipping, where the calibrated vial activity is determined using a liquid 90Y-chloride metrology standard.
To investigate the validity of this hypothesis, we implemented 3 vendor-specific vial geometries and compositions in dedicated MC simulations and compared the findings of this study with experimental results recently published in the literature.
MC Simulations
We simulated measurement of the activity of the 3 radioactive sources and respective vials with a commercial activity meter, using GAMOS 6.2 (14), a user-friendly interface of the GEANT4 code (15–17) extensively validated in the literature (18).
The implementation of the geometries will be detailed in the following subsections.
Concerning the simulated radioactive sources, 90Y is a β-emitter decaying to 90Zr with a half-life of 64.05 h. It has an endpoint energy of 2.28 MeV and an average energy of 0.936 MeV, giving the β-particles a maximum range of approximately 1.1 cm in water (19). For the 3 radioactive sources, we used the GEANT4 RadioactiveDecay module, which simulates the 90Y β-decay considering the creation of the chosen isotope and following its decay chain until a stable ion is produced.
As highlighted in our recent papers (20,21), GEANT4 RadioactiveDecay does not include internal bremsstrahlung (IB) emission accompanying the β-decay.
Consequently, to account for IB emission, an additional source term emitting photons was considered and photon energy was sampled according to the IB spectrum model validated by Auditore et al. (21) (supplemental data section A; supplemental materials are available at http://jnm.snmjournals.org).
Concerning the physics models, GmEMExtendedPhysics, which uses by default the Livermore low-energy electromagnetic interaction models, including atomic deexcitation, was applied (22). In each simulation, we ran 108 events to obtain results with statistical relative errors of less than 2%. The particle spatial propagation considered a range cutoff of 1 μm. No variance reduction technique was used. Simulations were run on an Intel Core i7 fourth-generation processor; each simulation with 90Y sources required about 6 h, whereas each simulation with IB photons as primary particles required less than 1 h.
Finally, we evaluated the accuracy of the performed MC simulation as a function of the assumed geometry for the vials and the activity meter. To achieve this aim, we varied the thickness of the walls in a range ±3σ with respect to the mean values used to build the model, where σ is the SD associated with the assumed dimensions.
The Activity Meter
We implemented in the MC simulations the geometry and composition of the activity meter Veenstra, model VDC-405, by Comecer, according to the technical layouts provided by the vendor (23). This setup has already been validated in our previous work (22,24), and more geometric details are reported in supplemental data section B.
The energy deposition, Edep, in the sensitive volume of the activity meter was scored for all simulated sources and converted into electric current per source activity (pA/MBq), as discussed by Auditore et al. (21) and reported in supplemental data section C.
The electric current per source activity obtained for 90Y glass and resin microspheres, Iglass and Iresin, were compared with that for 90Y-chloride by calculating the relative percentage differences, and
, as
Eq. 1where k stands for glass or resin.
Moreover, Iglass was compared with Iresin, estimating the relative percentage difference, δ:
Eq. 2
Finally, values of δ obtained in this work were compared with the relative percentage difference between glass and resin microsphere activity calibrations, δ*, obtained by Gnesin et al. (13).
90Y-Chloride Source
The 90Y-chloride source was implemented in MC simulations by homogeneously sampling 90Y decaying nuclides in 0.5 mL of water (G4_water, density [d] = 1 g/cm3) contained in a Pyrex (Corning) vial (G4_Pyrex glass, d = 2.23 g/cm3) with an aluminum cap (G4_aluminum, d = 2.699 g/cm3). Dimensions of a 90Y-chloride vial (Curium Pharma) were carefully measured to reproduce the actual vial geometry in the MC simulation as shown in Figure 1A. The concerned vial was 1 of the 2 90Y-chloride vials measured at Centre Hospitalier Universitaire Vaudois by Gnesin et al. (13).
Three-dimensional view of 90Y-chloride (A), 90Y glass (B) and 90Y (C) resin microspheres, as simulated with GAMOS.
90Y Glass Microsphere Source
Commercially available 90Y glass microspheres consist of insoluble glass microspheres with an average diameter ranging from 15 to 35 μm, as stated on the manufacturer’s datasheet (25), and an average density of 3.29 g/cm3, as stated by Paxton et al. (26). 90Y is present throughout the volume of the microspheres, conferring to each an activity of about 2,500 Bq. 90Y glass microspheres are available in standard activities ranging from 3 to 20 GBq at calibration, as reported in the manufacturer’s manual (27). The bottom of the vial containing 90Y glass microspheres in solution with water has a v-shape, and its dimensions have been carefully measured to reproduce the correct geometry in MC simulation as shown in Figure 1B. The vial is made of Pyrex and has an aluminum cap.
In the simulated configuration, the vial is placed in a vertical position inside the activity meter; 90Y glass microspheres deposit on the bottom of the vial, and the volume of the precipitate depends on the number of microspheres inside the solution, which in turn depends on the activity required by the user. The rest of the vial is filled with water, for a total filling volume of 0.8 mL.
The 90Y glass microsphere precipitate was reproduced in MC simulation as a glass matrix made of Y2O3 (user-defined Y2O3, d = 5.01 g/cm3), aluminum oxide (G4_aluminum oxide, d = 3.97 g/cm3), and silicon dioxide (G4_silicon dioxide, d = 2.32 g/cm3) in the proportion 40% Y2O3 to 20% Al2O3 to 40% SiO2 (percentages by weight) (28) with interstitial water. Primary particles (90Y and IB photons) were sampled uniformly throughout the glass matrix.
The volume and density of the 90Y glass precipitate, and
, were calculated taking into account random close packing for hard spheres (29), as reported in supplemental data section D.
For this study, an activity of 3 GBq was chosen, corresponding to 1.2 million microspheres in the vial; the 90Y glass microsphere precipitate, made of 85.09% glass matrix and 14.91% water (percentages by weight), with a density of 2.45 g/cm3, has a volume of 1.55⋅10−2 cm3.
Nevertheless, to verify the robustness of this study on the considered activity, and consequently on the volume occupied by the 90Y glass microspheres, additional simulations were performed for the commercially available activities listed in Table 1.
90Y Glass Microspheres Activities Considered for Robustness Analysis
90Y Resin Microsphere Source
90Y resin microspheres consist of insoluble resin microspheres with a mean diameter of 32 ± 10 μm and a density of 1.125–1.6 g/cm3 (30). 90Y is affixed to the resin microspheres via an ion exchange process and is present only on the surface of the microspheres. Each vial contains 5 mL of solution with 3 GBq of 90Y and 44 (±2.6) million resin microspheres (30) that deposit at the bottom. The dimensions of the vial were carefully measured to reproduce the correct vial geometry in the MC simulation, as shown in Figure 1C. The vial is made of Pyrex and has an aluminum cap.
The 90Y resin microsphere precipitate was reproduced in MC simulation as a polystyrene matrix (G4_polystyrene, d = 1.06 g/cm3) with interstitial water. Primary particles (90Y and IB photons) were sampled uniformly throughout the resin matrix.
The volume and density of the precipitate were calculated as for 90Y glass microspheres (supplemental data section D). For 3 GBq of activity, the volume of the precipitate was 1.19 cm3, and it was made of 73.49% polystyrene matrix and 26.51% water (percentages by weight), with a density of 1.38 g/cm3.
Scoring of Photons Emitted by Sources
Besides the energy deposition in the active volume of the activity meter, to better understand the obtained results, we scored the energy distribution of photons escaping the vial and contributing to the signal generated in the activity meter. To this aim, the World volume and the detector were set to the material G4_galactic (empty space) and the photons were scored at the inner cylindric surface of the activity meter, classifying them according to their energy. The estimated photon spectra for 90Y-chloride, 90Y glass microspheres, and 90Y resin microspheres were then compared with each other.
Comparison with PET Measurements from Literature
Assuming that at the manufacturer’s site the activity meter calibration factor, cM, is obtained with a standard 90Y-chloride solution and then used to measure the activity of 90Y sources in different vial configurations, the manufacturer activity is calculated as
Eq. 3where x stands for glass, chloride, or resin and i is the current measured by the activity meter.
In this approach, cM is assumed to be the same for resin, chloride, and glass solutions.
In this study, we hypothesized—also in light of the results reported by Gnesin et al. (13)—that the actual calibration factors, cx, depend on the vial geometry and on the source configuration; consequently, the activities should be calculated using the proper calculation factor:
Eq. 4where
is the actual activity.
Gnesin et al. measured the APET/AM ratios for 90Y glass and resin microspheres and 90Y-chloride solution. Assuming APET as the actual activity, , we can express the measured ratios as
Eq. 5And then the APET/AM ratios measured by Gnesin et al. would reflect the difference between the used and the actual calibration factors. It should be stressed that APET/AM was found by Gnesin et al. to be equal to 1 for chloride vials, experimentally confirming that the manufacturer standard is chloride solution and thus that AM = Achloride.
In this study, we estimate, by MC simulation, Ix = ix/Ax, which, from Equation 4, results in
Eq. 6
Using Equation 6 in Equation 5, we obtain
Eq. 7
Referring to 90Y-chloride solution as a common metrologic reference (cM = cchloride) and assuming the same amount of activity for each vendor vial (AMchloride = AMglass = AMresin), it follows that
Eq. 8
Eq. 9
Therefore, electric current ratios obtained from MC simulations can be directly compared with the PET-derived activity ratios reported by Gnesin et al. (13).
RESULTS
In Table 2 and Figure 2, we reported the MC estimates obtained in terms of electric current per unit activity (sampling only 90Y as the primary source, without IB). A 90Y glass microsphere vial gives an electric signal 30.7% higher than 90Y-chloride, whereas a signal 17.2% lower is obtained for a 90Y resin microsphere vial (Fig. 2B); these values are in fair agreement with the differences found by Gnesin et al. (13) of 27% and −13% between activities measured with PET and by the glass and resin microsphere manufacturers. The relative difference between 90Y glass and resin microspheres is 57.8% ± 2.4%, a value higher than the one reported by Gnesin et al. (13).
Electric Current per Unit Activity Estimated for 90Y-Chloride, 90Y Glass Microspheres, and 90Y Resin Microspheres, Sampling 90Y as Primary Particle (β-Decay)
(A and C) Electric currents per unit activity estimated from sampling standard 90Y β-decay spectrum (A) or 90Y β-decay with addition of IB photons (C). (B and D) Relative percentage differences, and
, calculated for MC results without (B) and with (D) IB photons.
To obtain more accurate estimates, and in light of the relevance of IB emission pointed out in our recent studies (20,21), MC simulation was performed also including IB photons, , as an additional source term, and the results are reported in Table 3 and shown in Figure 2C. Simulating 90Y and IB photons as primary particles, we found that the differences between the electric current per unit activity for 90Y glass and resin microspheres and for 90Y-chloride reduces to 24.6% ± 3.9% and −15.0% ± 2.2% (Fig. 2D), thus improving agreement with the measurements from Gnesin et al. (13). Also the relative percentage difference between 90Y glass and resin microspheres reduces to 46.5% ± 1.9%, in very good agreement with the values reported by Gnesin et al. (13).
Electric Current per Unit Activity Estimated for Considered Sources, with Additional Sampling of IB Photons, , as Primary Particles
Table 4 reports the ratios of the current per unit activity of chloride versus glass and chloride versus resin configurations, obtained with MC simulations including IB contribution. These values are in excellent agreement with the activity ratios found by Gnesin et al. (13), confirming the assumptions behind Equations 8 and 9.
Ratios Between Activities of Glass and Resin Microspheres and 90Y-Chloride Vials
The geometry used in MC simulations was varied to evaluate the robustness of the model, with respect to the variability in the dimensions of the activimeter and vial geometry. For the applied geometry variations to the activity meter, within the range ±3σ, εglass resulted in a CI of ±12.6%, an εresin of ±10.2%, and a δ of ±2.2%.
The modification of the vial dimensions leads to variations of about ±9.4% and ±10.7% for εglass and εresin, respectively, and ±2.8% for δ.
When the activity meter and vial geometry variations were combined, the CIs for εglass, εresin, and δ were estimated to be ±15.9%, ±14.9%, and ±4.1%, respectively.
The activity of the vial containing 90Y glass microspheres depends as much on the number of microspheres as on the volume occupied by them (Table 1). To test the robustness of this study as a function of the considered 90Y glass microsphere activity, the electric currents per unit activity obtained for the activities listed in Table 1 are reported in Table 5 for comparison purposes. Taking the 3-GBq activity source as the reference, the relative percentage difference between the results is lower than 1%, thus indicating the independence of the MC estimates from the source activity and the robustness of the study in the considered range.
Electric Currents per Unit Activity Obtained for Different Activities of 90Y Glass Microsphere Source
The photon spectra scored at the inner surface of the activity meter are reported in Figure 3. The energy peak at 1,760 keV results from the deexcitation of an excited state of 40Zr to which 90Y decays (19). Due to the shape and thickness of the vial, particularly in the v-shaped bottom, the photon yield estimated for the 90Y glass microsphere vial is higher than the yields resulting from 90Y-chloride and 90Y resin microspheres, the latter providing the lowest photon yield. The relevant differences in the energy range 0–1 MeV (Fig. 3B) are reflected in the different electric currents generated in the activity meter for the 3 source vials.
(A) Photon spectra sampled at inner surface of activity meter. (B) Zoomed 0- to 1-MeV energy range with linear y scale. Primary particles sampled were 90Y and IB photons, .
DISCUSSION
In this work, we performed MC simulations of the activity meter signal generation from 90Y sources in 3 commercial vial configurations. We simulated the 90Y decay generating the external bremsstrahlung photons due to the interaction of the β-particles with the surrounding material; in addition, we included the IB photons that originate from the interaction of the emitted β-particles with the electromagnetic field of the emitting nucleus.
We calculated the electric signal produced in the 3 source configurations and computed the relative percentage difference comparing the 2 types of 90Y-labeled microspheres with the 90Y-chloride solution. Moreover, we assessed the relative percentage difference between the electric signal obtained from the glass and resin microspheres.
Taking into account also the IB contribution, we found that for a matched activity of 90Y, the electric signals generated by the 90Y-labeled glass and resin microsphere vials were 24.6% higher and 15.0% lower, respectively, than the signal produced by the vial containing a homogeneous 90Y-chloride solution. Hence, we found a relative difference of 46.5% ± 1.9% for the electric signal estimated for the glass versus resin commercial vials.
We compared the findings of this study with the experimental results recently reported by Gnesin et al. (13). We found remarkably good agreement between the relative percentage differences in the electric signal produced by resin versus glass microspheres obtained with MC simulation and the relative percentage difference in total vial activity measured in a 90Y PET/CT experimental setup (13). Consequently, equally good agreement was found comparing the ratios of chloride versus glass and chloride versus resin electric signals with the corresponding ratios of activity measured in PET/CT, as detailed in Equations 8 and 9.
Therefore, the measured discrepancies in the activity assessment from commercial vials could be reasonably ascribed to the differences in the geometry and source composition across the 3 vials considered in the study.
PET measurements rely on the detection of coincidence annihilation photons and are, in principle, not sensitive to possible difference in the bremsstrahlung energy spectrum emitted by the different commercial products. Interestingly, Gnesin et al. (13) found the best agreement of PET-assessed activity with the vendor-declared calibrated activity for the 90Y-chloride vials. This finding can be explained if we assume that commercial activity meters (with the inclusion of activity meters used by the 90Y manufacturers) rely on a metrologic standard using a 90Y homogeneous liquid compound. Conversely, activity assessments of resin and glass microspheres, characterized by vial geometry and source composition sensibly different from the metrologic 90Y liquid standard, are prone to provide a potentially inaccurate activity evaluation if measured in a device calibrated with a 90Y liquid standard.
Assuming a common metrologic standard related to 90Y liquid chloride solution, as done in Equation 3, glass and resin vial activity assessments from the manufacturer activity meters would result in a systematic overestimation and underestimation, respectively, of the actual ones. Consequently, the glass and resin manufacturers filled their own vials with a lower and higher amount, respectively, of activity than the intended ones. This suggests that product-specific calibration factors should be adopted for each vial configuration.
Furthermore, the results presented by Gnesin et al. (13) were best reproduced when considering the IB in addition to the standard 90Y decay spectrum. This aspect can be understood considering that the electric signal generated in the activity meter results from interactions of all photons that deposit energy—indeed, both internal and external bremsstrahlung photons.
To our knowledge, this is the first study providing a reasonable explanation for the measured disagreement in the total 90Y activity obtained in PET/CT (in a multicenter and multidevice study) when compared with calibrated activity information provided by vendors (13).
In addition, we tested the robustness of the results by varying the geometric dimensions of both the activimeter and the vials. Estimations of εglass and εresin were found to vary within 16% (±15.9% and ±14.9%, respectively). The relative percentage difference between glass and resin microsphere activity calibrations, δ, shows a CI of ±4.1%, indicating that the considered geometric variations mildly influenced this parameter.
The estimate of 46.5% ± 1.9% for δ must be compared with the experimental δ* obtained from data reported by Gnesin et al. (13), who found δ* to equal 46.00% without any direct indication of the CI. From data published in Gnesin et al. (13), APET/AM (resin) = 1.15 ± 0.06 and APET/AM (glass) = 0.79 ± 0.04, and the CI can be estimated as follows:
Consequently, δ*upper = 61% and δ*lower = 31% and a CI can be assigned to δ* as 46% ± 15%. The MC-estimated δ reproduces the centroid of the gaussian distribution of the experimental δ* and is fully included in its CI (46.5% ± 1.9% vs. 46.00% ± 15%).
Another factor influencing the accuracy of the obtained results concerns the IB spectral distribution used in this study; it was obtained according to the fit of the IB spectrum measurements available in the literature, as reported in our previous work in which we identified the most appropriate model for the 90Y IB spectrum after comparing radiometric measurements and MC simulations (21,24); nevertheless, in perspective, new 90Y IB spectrum measurements could help to further refine the model.
CONCLUSION
The MC simulations of the response of a commercial activity meter to sources of 90Y-chloride, 90Y resin microspheres, and 90Y glass microspheres indicate that the discrepancies recently found between PET/CT-measured activity and vendor-calibrated activity for 90Y glass and resin microsphere vials are attributable to the difference in the geometry of the commercial vials, the specific spatial distribution of microspheres in solution or precipitate inside the vial, and the metrologic reference adopted for the activity meter calibration. Finally, the inclusion of IB photons in MC simulations dealing with the 90Y source is advisable because it plays a relevant role in determining the activity meter output.
DISCLOSURE
No potential conflict of interest relevant to this article was reported.
KEY POINTS
QUESTION: What is a possible explanation for the experimentally observed discrepancies between 90Y vial manufacturer-calibrated activities and independent PET quantitative assessment?
PERTINENT FINDINGS: Considering a well counter configuration typical of commercial activity meters, we used MC simulations to assess the electric signal generation for the measurement of commercial vials containing 90Y-labeled glass and resin microspheres and 90Y-chloride in liquid solution. For the same vial activity, simulation gave a relative electric current difference of +24.6% for glass vials and −15% for resin vials as compared with the 90Y liquid chloride solution.
IMPLICATIONS FOR PATIENT CARE: Accurate activity assessment is mandatory for therapeutic procedures and key for the establishment of reliable dose–effect studies.
Footnotes
Published online Jul. 13, 2023.
- © 2023 by the Society of Nuclear Medicine and Molecular Imaging.
REFERENCES
- Received for publication January 20, 2023.
- Revision received April 25, 2023.