[68Ga]Ga-PSMA-11 PET imaging as a predictor for absorbed doses in organs at risk and small lesions in [177Lu]Lu-PSMA-617 treatment

Introduction
 Patient eligibility for [177Lu]Lu-PSMA therapy remains a challenge, with only 40–60% response rate when patient selection is done based on the lesion uptake (SUV) on [68Ga]Ga-PSMA-PET/CT. Prediction of absorbed dose based on this pre-treatment scan could improve patient selection and help to individualize treatment by maximizing the absorbed dose to target lesions while adhering to the threshold doses for the organs at risk (kidneys, salivary glands, and liver). Methods Ten patients with low-volume hormone-sensitive prostate cancer received a pre-therapeutic [68Ga]Ga-PSMA-11 PET/CT, followed by 3 GBq [177Lu]Lu-PSMA-617 therapy. Intra-therapeutically, SPECT/CT was acquired at 1, 24, 48, 72, and 168 h. Absorbed dose in organs and lesions (n = 22) was determined according to the MIRD scheme. Absorbed dose prediction based on [68Ga]Ga-PSMA-PET/CT was performed using tracer uptake at 1 h post-injection and the mean tissue effective half-life on SPECT. Predicted PET/actual SPECT absorbed dose ratios were determined for each target volume. Results PET/SPECT absorbed dose ratio was 1.01 ± 0.21, 1.10 ± 0.15, 1.20 ± 0.34, and 1.11 ± 0.29 for kidneys (using a 2.2 scaling factor), liver, submandibular, and parotid glands, respectively. While a large inter-patient variation in lesion kinetics was observed, PET/SPECT absorbed dose ratio was 1.3 ± 0.7 (range: 0.4–2.7, correlation coefficient r = 0.69, p < 0.01). Conclusion A single time point [68Ga]Ga-PSMA-PET scan can be used to predict the absorbed dose of [177Lu]Lu-PSMA therapy to organs, and (to a limited extent) to lesions. This strategy facilitates in treatment management and could increase the personalization of [177Lu]Lu-PSMA therapy. Supplementary Information The online version contains supplementary material available at 10.1007/s00259-021-05538-2.

At present, high tumor uptake of [ 68 Ga]Ga-PSMA-11, [ 18 F]DCFPyL, or [ 18 F]PSMA-1007 on positron emission tomography (PET) imaging is mandatory for PSMA radioligand therapy [26][27][28][29][30]. In some studies, PET standardized uptake value (SUV) on [ 68 Ga]Ga-PSMA-PET has been shown to correlate with absorbed (radiation) dose in lesions and salivary glands in both mCRPC [31] and mHSPC patients [32], while other studies did not find this correlation [9,18,33]. Patients selected based on PET lesion SUV show a response rate of only 40-60% [7][8][9][10][11][12][13][14][15][16][17][18][19]. A potential improvement of patient selection has been suggested by the group of Hofman and colleagues by using both FDG-positive tumor volume and mean intensity of PSMA-avid tumor uptake [34]. An actual dose estimation based on the pre-therapeutic [ 68 Ga]Ga-PSMA-PET could provide more accurate information on expected treatment response, since the calculations of the absorbed doses take into account tracer kinetics and are intrinsically corrected for factors such as partial volume effect occurring in particular for small tumors. Therefore, we hypothesize that patient selection could be improved if the pre-therapeutic [ 68 Ga]Ga-PSMA-PET data were used to predict absorbed doses for the subsequent [ 177 Lu]Lu-PSMA treatment.
In addition, pre-therapeutic evaluation of risk for organ toxicity is important in order to design a patient-specific treatment plan. It can prevent clinicians from exceeding threshold doses for radiation-related toxicity and it can potentially be used to apply higher therapeutic activities. To this end, mean SUV of organs on PET are not a suitable parameter to predict organ absorbed dose, mainly due to heterogeneity in the PET signal. However, modeling the organ absorbed doses based on the pre-therapeutic [ 68 Ga] Ga-PSMA-PET imaging could provide a tool to assess organ toxicity after treatment.
Similar studies have been carried out using PET/CT imaging for an absorbed dose prediction after radionuclide therapy, mainly using 124 I for prediction of organ-absorbed dose after 131 I-therapy in thyroid cancer patients [35][36][37][38][39]. This methodology is based on the assumption that tracer kinetics for 124 I and 131 I are comparable, and cumulated activity derived from multi time point 124 I-PET/CT can be translated to 131 I-cumulated activity, thereby predicting organ absorbed dose after therapy. It is suggested that this approach can be used to design patient-specific treatment by respecting the organ threshold dose for radiation toxicity effects [39,40].
To It relies on tissue-specific radioligand kinetics that will be derived from therapeutic imaging data with [ 177 Lu] Lu-PSMA-SPECT, in combination with tracer uptake of a single time point pre-therapeutic [ 68 Ga]Ga-PSMA-PET. The predicted absorbed doses were compared to actually delivered absorbed doses in therapy.

Patient population
The study comprised 10 patients with low-volume hormone-sensitive prostate cancer who received [ 177 Lu]Lu-PSMA for treatment of oligometastatic prostate cancer. It was approved by the Medical Review Ethics Committee Region Arnhem-Nijmegen and was registered on clinicaltrials.gov (NCT03828838). The trial was done in accordance to the principles of Good Clinical Practice and the Declaration of Helsinki. All subjects provided written informed consent before study entry. A comprehensive description of the patient population has been published before [25]. In short, HSPC patients with prostate-specific antigen (PSA) doubling time ≤ 6 months and ≤ 10 visible metastases on baseline [ 68 Ga]Ga-PSMA-PET/ CT, with at least one lesion ≥ 10 mm in diameter, were included. Normal renal and bone marrow functions were required (MDRD-GFR ≥ 60 ml/min, white blood cell count > 3.5 × 109.131/l, platelet count > 150 × 109.132/l and hemoglobin > 6 mmol/l). A detailed study flowchart can be found in Online Resource 1.

Imaging and therapy
Patients received [ 68 Ga]Ga-PSMA-11-PET/CT approximately 1 week prior to radioligand therapy. Imaging was performed 60 ± 10 min post-injection on a Biograph mCT system (Siemens Healthineers, Erlangen, Germany) scanning cranium to trochanter major. Patients received a therapeutic activity of 3 GBq (3057 ± 38 MBq) [ 177 Lu]Lu-PSMA-617. This relatively low activity was chosen because it was part of a prospective pilot study in a patient population that did not receive this type of treatment before. The preparation of [ 177 Lu]Lu-PSMA was described previously [25] and can be found in Online Resource 2. SPECT/CT imaging was performed at 1, 24, 48, 72, and 168 h after administration on either a Symbia T16 or Symbia Intevo Bold system (Siemens Healthineers, Erlangen, Germany). SPECT/CT scans were acquired at three body regions: pelvis, abdomen, and headneck regions. Acquisition and reconstruction parameters can be found in Online Resource 3.

Organ and tumor volumes
Organ and tumor volumes were derived by manual segmentation (VOI technique) using the reference CT image. Some structures could not be reliably delineated on CT. As an alternative, volumes of the paired parotid and submandibular glands as well as bone lesions were determined using a PET-based iterative thresholding method [41]. It has been shown that this technique allows for accurate volume estimation down to the 68 Ga-PET spatial resolution of 0.13 ml and reveals reliable volume estimates for objects with moderate non-uniform activity distributions.

Imaged SPECT and PET activities in organs and lesions and their corrections
To determine the SPECT and PET activities in organs, a contour-based approach was applied: the imaged activity, A ContourVOI , within the contour of the organ boundary was determined and corrected for partial volume effects using the fitted isovolume recovery coefficient, RC iso . The corrected organ activity A Corrected is given as follows: The isovolume RC values depend on spatial resolution and patient's individual organ volumes [42].
The diameters of the lesions were small (median diameter of 12 mm, range: 6-43 mm), and thus clearly below the SPECT spatial resolution of 15 mm. For a tissue size smaller than 1.25-1.5 times the spatial resolution, correction for partial volume effect is not recommended [42,43]. Instead, an oversize-based method was applied [44][45][46]. An oversized lesion VOI was drawn large enough to include the entire tail of the lesion activity, and its cross-contamination from the surrounding background activity was further removed using the following equation: where A Corrected is the corrected lesion activity, V Lesion is the lesion volume, A OversizeVOI is the imaged activity of the oversized lesion VOI, and c Background is the background activity concentration derived from a representative background VOI close to the lesion. Of note, the drawback of this oversizedbased approach is that high and non-uniform background may result in activity underestimation.

Biokinetic analysis-parametrization of the tissue-specific radioligand uptake curves
The observed tissue-specific [ 177 Lu]Lu-PSMA uptake curves were analyzed to derive a typical time-activity curve (TAC) or, equivalently, uptake curves for each organ and for the lesions. First, to avoid ambiguity and maintain consistency for all uptake curves, no curve-fitting procedure was applied. Hence, the common approach to fit the 5 data points at once using, for instance, a bi-exponential function, was not used. Instead of fitting the entire dataset, the biokinetic data were segmented into three phases, that is, an early, a mid, and a late phase to extract the typical uptake pattern of each tissue type within a time segment. Second, the exploratory investigation of the intra-therapeutic uptake curves revealed that there are three types of uptake patterns. Figure 1 schematically illustrates the 3 phases and the respective piecewise parameterization of representative uptake curves. For kidneys and liver, an instantaneous uptake (early phase) followed by a mono-exponential clearance with an effective half-life T eff,1 up to 72 h (mid-phase), and thereafter, a second mono-exponential clearance (late phase) with an effective half-life T eff,2 ( Fig. 1A) was observed. Thus, the early and mid-phases were parameterized using the half-life T eff,1 and the late phase was parameterized using half-life T eff,2 ; the respective half-lives were obtained by linear regression analyses. For the salivary glands, kinetics was assessed for Fig. 1 Types of uptake pattern depending on tissue type. A Segmented bi-exponential uptake curve for liver and kidneys. B Rectangular monoexponential uptake curve for salivary glands. C Linear mono-exponental curve for lesions the whole organ instead of separate glands. An instant uptake was observed, and its value remained almost constant with a (average) value of U 0 up to 24 h (early and midphases). More precisely, the 24 h value was sometimes above or below the 1 h uptake value; the average uptake value was used to effectively represent these phases. Thereafter, a mono-exponential clearance with an effective half-life T eff (late phase) was found (Fig. 1B). For lesions, a linear increase (early phase) was observed with a slope derived from the first uptake value U(t 1 ), that is, α = U(1) a maximum uptake U max (mid-phase) followed by a monoexponential decay (late phase) with an effective half-life T eff (Fig. 1C), with the intercept of both functions at time t max . For clarity, the respective equations are given in Online Resource 4.
The mean values of the intra-therapeutic kinetic parameters were calculated and used to predict the biokinetics for each tissue type based on a single PET-based uptake value of [ 68 Ga]Ga-PSMA.

Pre-therapeutic [ 177 Lu]Lu-PSMA dosimetry using [ 68 Ga]Ga-PSMA SPECT/CT imaging
For the absorbed dose prediction, a correction regarding the differences in the physical half-lives of 68 Ga (T Ga ) and 177 Lu (T Lu ) is necessary according to the radioactive decay law. The observed [ 68 Ga]Ga-PSMA uptake value U Ga (t PET ) was projected to the predicted [ 177 Lu]Lu-PSMA uptake value U Lu (t PET ) using the following equation [47]: This projected uptake value was used to construct the individual uptake curve based on the tissue-specific biokinetic model, from which the projected [ 177 Lu]Lu-PSMA residence times (or TIAC values) were estimated. In the construction of the projected uptake curve, the tissue-specific mean values of the intra-therapeutic kinetic parameters were applied. The projection of the functions to determine the TIAC values is given in Online Resource 5.
For each organ and lesion, the projected [ 177 Lu]Lu-PSMA TIAC and the mass were used to predict the absorbed dose per unit administered [ 177 Lu]Lu-PSMA activity using Olinda 2.2.

Software and statistical analysis
The image interpolation and image analyses were conducted using PMOD 4.2 software (PMOD Technologies Ltd., Zurich, Switzerland). Statistical analysis was performed using GraphPad Prism 5.03 (Graphpad Software Inc., CA, USA). The descriptive statistics included the mean, median, standard deviation (SD), and range and were expressed in the following form: mean ± SD (median, minimum-maximum

Analysis of the intra-therapeutic biokinetics
Uptake curves were determined based on the effective halflives for each organ ( Table 2) and lesions (Table 1). Kidneys showed a median half-life of 28 h and 49 h for the first and second excretion phases, respectively. Liver showed a median half-life of 21 h and 47 h for the first and second excretion phases, respectively. For the salivary glands, uptake between t 0 and t 24h was assumed to be constant based on the average uptake of U 1h and U 24h (Fig. 1B). For the second phase after 24 h, median half-life was 33 h.
For 6 out of the 22 lesions, the volume was not visible on the 1-h time point [ 177 Lu]Lu-PSMA-SPECT; therefore, the initial uptake kinetics could not be determined individually. Instead, uptake phase was estimated by taking the mean t max for the other 16 lesions, which was 3.9 h. Median effective half-life T eff for clearance after 24 h was 72 h. Of note, kinetics were not significantly different between bone and lymph node lesions (Mann-Whitney U-test: p = 0.86).

Intra-therapeutic and predicted absorbed doses for organs
The median absorbed dose as determined from [ 177 Lu] Lu-PSMA-SPECT as well as the median predicted absorbed dose from [ 68 Ga]Ga-PSMA-PET can be found per organ in Fig. 3. Combined statistics and organ kinetics per patient can be found in Online Resources 8-11. For the kidneys, initial PET/SPECT absorbed dose ratio was 2.21 ± 0.46 (1.32-2.75). Because the ratio was rather constant, a scaling factor F = 2.2 was introduced by which the PETpredicted absorbed dose was divided, leading to a PET/ SPECT absorbed dose ratio for the kidney was 1.01 ± 0.21 (0.60-1.25). For liver, submandibular glands, and parotid glands, agreement between SPECT and PET absorbed dose was high, with PET/SPECT absorbed dose ratios of 1.10 ± 0.15 (0.94-1.35), 1.20 ± 0.34 (0.61-1.84), and 1.11 ± 0.29 (0.54-1.47), respectively. No scaling factor was introduced for these organs.

Intra-therapeutic and predicted absorbed doses for lesions
The results per lesion can be found in Table 3 Fig. 4B). Lesion volume dependency of PET/SPECT absorbed dose ratio is shown in Fig. 5, where it can be seen that an underestimation of absorbed dose was mainly found for the smaller lesion volumes (Spearman significant correlation, r = 0.43, p < 0.05).

Discussion
This study evaluated the possibility to use a single time point [ 68 Ga]Ga-PSMA-PET scan to predict the therapeutic absorbed dose in organs at risk and lesions for a subsequent treatment with [ 177 Lu]Lu-PSMA therapy. Tracer kinetics is a crucial part in these predictive absorbed dose calculations, which determines the shape of the uptake time-activity curve and thus the cumulated activity. In this study, these kinetics were determined as the mean kinetics of 10 patients based on the SPECT data. This means that this approach is based on two main assumptions: firstly, the typical shape of the uptake curves for organs and lesions for the different patients are nearly identical; therefore, it is justified to use general tissue-specific kinetics in the PET prediction model. Secondly, the different tracers used in PET and SPECT imaging (PSMA-11 and PSMA-617, respectively) have similar kinetics; therefore, the kinetics found for PSMA-617 on SPECT can be used to project the expected kinetics of PSMA-11 on PET. Multiple studies investigated biodistribution and kinetics for PSMA-11 [48][49][50][51] and PSMA-617 [52] and showed indeed similar kinetic behavior [53]. Prediction of absorbed dose for lesions showed a large variation in kinetics between patients both during the uptake phase (SD of 50%) and the excretion phase (SD of 30%), indicating that the first assumption of identical lesion kinetics between patients does not hold. Therefore, tumor lesion dosimetry using a single time [ 68 Ga]Ga-PSMA-PET was challenging. Earlier studies found that different tracer kinetics could be the result of different lesion types (bone versus lymph node lesions) [54]. However, in our study no statistically significant difference in tracer kinetics between the two tissue types was found (p = 0.84). The highly variable kinetics observed in lesions are possibly the result of heterogeneity in tumor biology. Therefore, the use of a general tracer uptake pattern for lesions will introduce relevant deviations on an individual level.
However, the proposed methodology using lesion-specific kinetics results in a rather good PET/SPECT absorbed dose ratio for lesions of 1.3 ± 0.7 (0.4-2.7), with a significant correlation (r = 0.69, p < 0.01) that was not found between SUV max on [ 68 Ga]Ga-PSMA-PET and absorbed dose after therapy (r = 0.16, p = 0.47). So, despite a relatively large range in PET/SPECT absorbed dose ratio, an actual absorbed dose prediction could still mean a significant improvement in patient selection compared to only using lesion SUV max , since it provides better insight in what lesion uptake is to be expected and thus whether treatment with [ 177 Lu]Lu-PSMA is expected to be effective.
Estimation of patient-specific tracer uptake in lesions could potentially be improved by obtaining continuous information on tracer distribution during the first hour after injection of [ 68 Ga]Ga-PSMA using dynamic PET imaging [54]. Moreover, obtaining uptake information at multiple later time points could provide crucial information on late tracer kinetics, which largely determine the absorbed dose.  [55]. Therefore, 89 Zr-labelled PSMA has the potential to improve lesion absorbed dose prediction.
The large range in PET/SPECT absorbed dose ratio found in this study can also partly be explained by difficulties in calculating SPECT absorbed dose for small structures, such as the lesions found in this patient cohort. Due to limited image resolution, count statistics, and photon scatter, determination of residence times is difficult. This means that in general, larger uncertainties in absorbed dose calculations are found in these small volumes [32,56]. While patient selection might be improved by combining lesion SUV on PSMA-PET with evaluation of positive tumor uptake on 18 F-FDG-PET [34], this does not provide information on risk of organ toxicity. The mHSPC patient cohort for this study, acute organ toxicity, is not anticipated, since these patients tend to have a relatively good physical condition and good organ function. However, development of chronic toxicities should be prevented. In addition, presently, [ 177 Lu]Lu-PSMA therapy is mainly applied in mCRPC patients, which are at risk for compromised organ function and may have received prior radionuclide therapy that already deposited a radiation dose to the healthy organs. Therefore, an absorbed dose prediction based on the pretherapeutic [ 68 Ga]Ga-PSMA-PET scan would provide the physician with a useful tool to manage or refrain from additional treatment cycles when there is a significant risk of organ toxicity. Our study showed that an absorbed dose prediction based on a single time point [ 68 Ga]Ga-PSMA-PET  [35][36][37][38][39][40]. Tissue-specific organ kinetics showed to be stable between patients, which means that uptake information at a single time point in combination with assumed tissue-specific tracer kinetics provide an effective instrument for absorbed dose prediction. Although it was shown earlier that organ tracer kinetics in mHSPC patients are very similar to those in mCRPC patients [32], it would be advised to establish tissue-specific tracer kinetics for mCRPC patients when applying the proposed methodology in this specific patient group. Furthermore, our results are based on only 10 patients. More elaborate data of larger patient cohorts is warranted. Initially, we found that the absorbed dose prediction based on PET for the kidneys was notably higher than the SPECT-based values: PET/SPECT absorbed dose ratio of 2.21 ± 0.46 (Fig. 3B). A possible explanation could be a difference in early phase kinetics between patients, which was the only exception found in this study that showed somewhat larger variation in tracer kinetics: 21% for the early phase kinetics up to 72 h ( Table 2). In addition, there might be a difference in 1-h tracer uptake between PSMA-11 and PSMA-617. Since the PSMA tracer is cleared mainly via the kidneys, a potential faster blood and renal uptake for PSMA-11 would lead to a higher activity found in the kidney at 1 h p.i. on PET than for PSMA-617 at the same time point on SPECT. In addition, there are some differences in the coordination chemistry of [ 68 Ga]Ga-PSMA-11 and [ 177 Lu] Lu-PSMA-617; Ga 3+ forms a hexadentate binding in the HBED chelator leaving two nitrogens and Lu 3+ a octadentate binding in the DOTA chelator. In preclinical setting, it was shown that this leads to a higher kidney uptake of [ 68 Ga] Ga-PSMA-11 in comparison to [ 111 In]In-PSMA-617 [57,58]. This would then lead to an overestimation of the total predicted absorbed dose. A remarkable feature was that the PET/SPECT absorbed dose ratio for kidney was rather constant. After applying a scaling factor of 2.2, a PET/SPECT absorbed dose ratio of 1.01 ± 0.21 (Fig. 3B) was obtained. Thus, despite an initial overestimation of kidney absorbed dose based on [ 68 Ga]Ga-PSMA-PET, it can reliably be used to predict therapeutic absorbed dose for [ 177 Lu]Lu-PSMA after applying the scaling factor, with a maximum deviation of around 20%. For the other organs, such a deviation was not found so no scaling was performed. However, the salivary glands showed a relatively large range in PET/ SPECT absorbed dose ratio for the submandibular glands (0.61-1.84) and parotid glands (0.54-1.47), respectively (Fig. 3). This indicates that, despite very comparable overall tracer kinetics in the salivary glands, the uptake at 1 h p.i. can be rather variable between patients, leading to a larger range in PET/SPECT ratio.

Conclusion
This study showed that a single time point [ 68 Ga]Ga-PSMA-PET scan can be used to predict the absorbed dose of [ 177 Lu] Lu-PSMA therapy to the kidney, liver, salivary glands, and (to a limited extent) to tumor lesions. The proposed methodology is readily available for clinical implementation since the pre-treatment PSMA-PET scan is already required for Acknowledgements We thank all the patients that participated in this study, and the staff at the departments of Medical Imaging, Urology and Medical Oncology of the Radboud University Medical Center for their involvement.
Author contribution All authors were involved in writing and reviewing of the manuscript. In addition, SP was involved in the study design, data analysis, and manuscript design. RH was involved in data analysis. BP was involved in data collection. MdB was involved in data collection. MG was involved in study design and data analysis review. MJ was involved in study design and patient inclusion. FdL was involved in data review. CM was involved in study design. NM was involved in study design, patient selection, and data collection. JW was involved in study design, patient selection, and data collection. PC was involved in data analysis review. JN was involved in study design, patient selection, data collection, and data analysis review. MK was involved in study design, data analysis review, and manuscript design. WJ was involved in data analysis methodology, data analysis, and manuscript design.
Funding Partial financial support was received from the Radboud Oncology Foundation and the Dutch Prostate Cancer Foundation.

Data availability
The datasets generated during and/or analyzed during the current study are available from the corresponding author on reasonable request.
Code availability Not applicable.

Declarations
Ethics approval All procedures performed in studies involving human participants were in accordance with the ethical standards of the institutional and/or national research committee and with the 1964 Helsinki Declaration and its later amendments or comparable ethical standards. This study was approved by the Medical Review Ethics Committee Region Arnhem-Nijmegen and was registered on clinicaltrials.gov (NCT03828838).
Consent to participate All subjects provided written informed consent before study entry.

Consent for publication
Patients signed informed consent regarding publishing their data.
Competing interests W. Jentzen received research funding from Siemens Healthineers. Partial financial support was received from the Radboud Oncology Foundation and the Dutch Prostate Cancer Foundation.
Open Access This article is licensed under a Creative Commons Attribution 4.0 International License, which permits use, sharing, adaptation, distribution and reproduction in any medium or format, as long as you give appropriate credit to the original author(s) and the source, provide a link to the Creative Commons licence, and indicate if changes were made. The images or other third party material in this article are included in the article's Creative Commons licence, unless indicated otherwise in a credit line to the material. If material is not included in the article's Creative Commons licence and your intended use is not permitted by statutory regulation or exceeds the permitted use, you will need to obtain permission directly from the copyright holder. To view a copy of this licence, visit http:// creat iveco mmons. org/ licen ses/ by/4. 0/.