[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

The study comprised 10 patients with low-volume hormone-sensitive prostate cancer who received [¹⁷⁷Lu]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 [⁶⁸Ga]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.

Patients received [⁶⁸Ga]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) [¹⁷⁷Lu]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 [¹⁷⁷Lu]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 head-neck regions. Acquisition and reconstruction parameters can be found in Online Resource 3.

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 ⁶⁸Ga-PET spatial resolution of 0.13 ml and reveals reliable volume estimates for objects with moderate non-uniform activity distributions.

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 [44‐46]. 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 oversized-based approach is that high and non-uniform background may result in activity underestimation.

The observed tissue-specific [¹⁷⁷Lu]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 the whole organ instead of separate glands. An instant uptake was observed, and its value remained almost constant with a (average) value of U₀ up to 24 h (early and mid-phases). 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(t₁), that is, α = \(\frac{U(1)}{{t}_{1}}\) (in %/h), to a maximum uptake U_max (mid-phase) followed by a mono-exponential 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 [⁶⁸Ga]Ga-PSMA.

For the absorbed dose prediction, a correction regarding the differences in the physical half-lives of ⁶⁸Ga (T_Ga) and ¹⁷⁷Lu (T_Lu) is necessary according to the radioactive decay law. The observed [⁶⁸Ga]Ga-PSMA uptake value U_Ga(t_PET) was projected to the predicted [¹⁷⁷Lu]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 [¹⁷⁷Lu]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 [¹⁷⁷Lu]Lu-PSMA TIAC and the mass were used to predict the absorbed dose per unit administered [¹⁷⁷Lu]Lu-PSMA activity using Olinda 2.2.

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). Uncertainty in the absorbed dose values were determined following the EANM uncertainty guideline by Gear et al. (for more details see Online Resource 6). Differences between the 2 groups were evaluated by the Mann–Whitney U test. A Spearman non-parametric correlation test was used to evaluate correlations between lesion SUV_max on [⁶⁸Ga]Ga-PSMA-PET and absorbed dose on [¹⁷⁷Lu]Lu-PSMA-SPECT, between lesion-absorbed dose based on [⁶⁸Ga]Ga-PSMA-PET and [¹⁷⁷Lu]Lu-PSMA-SPECT, and between lesion PET/SPECT absorbed dose ratio and lesion volume. A p-value of less than 0.05 was considered to be statistically significant.

Patient characteristics and administered activities (GBq) can be found in Online Resource 7. A total of 22 lesions were evaluated (1 to 7 per patient). For 8 lesions, volume determined on CT was 6.5 ± 14.6 ml (0.9, 0.21–42.5, Table 1). For the other 14 lesions, volume was determined on PET with a volume of 3.0 ± 5.3 ml (1.1, 0.13–20.2). In order to compare the methodology, volume was determined on both CT and PET for 9 lesions. This showed a mean ratio of 1.09 ± 0.17 for PET volume versus CT volume. The difference in volume was not significant (Mann–Whitney U test: p = 0.86).

Figure 2 shows an example of typical images for the organs of interest and two lesions for [¹⁷⁷Lu]Lu-PSMA-SPECT, CT, and [⁶⁸Ga]Ga-PSMA-PET. For the lesions, it shows that some are clearly visible on CT (Fig. 2A) while others are not (Fig. 2B), while both lesions in this example are clearly visible and delineable on PET as well as SPECT.

Uptake curves were determined based on the effective half-lives 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 t₀ 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 [¹⁷⁷Lu]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).

The median absorbed dose as determined from [¹⁷⁷Lu]Lu-PSMA-SPECT as well as the median predicted absorbed dose from [⁶⁸Ga]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 PET-predicted 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.

The results per lesion can be found in Table 3 (SUV_max on [⁶⁸Ga]Ga-PSMA-PET, as well as absorbed dose predicted from [⁶⁸Ga]Ga-PSMA-PET and determined from [¹⁷⁷Lu]Lu-PSMA-SPECT). The PET/SPECT absorbed dose ratio for lesions was 1.3 ± 0.7 (1.1, 0.4–2.7). No significant correlation was found between SUV_max on [⁶⁸Ga]Ga-PSMA-PET and absorbed dose from [¹⁷⁷Lu]Lu-PSMA-SPECT (r = 0.16, p = 0.47, Fig. 4A), while a significant correlation was found between predicted absorbed dose from [⁶⁸Ga]Ga-PSMA-PET and determined absorbed dose from [¹⁷⁷Lu]Lu-PSMA-SPECT (r = 0.69, p < 0.01, 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).