166Ho-MS preparation
Non-radioactive 165Ho-MS were prepared as previously described [13]. Neutron activation of the Ho-loaded microspheres was performed via the 165Ho (n, ɣ) 166Ho reaction in a nuclear reactor with a nominal thermal neutron flux of 5 × 1012 cm−2 s−1 (Delft University of Technology). Therefore, vials with known amounts of 165Ho-MS were irradiated up to 2 h. After neutron activation, 166Ho-MS emit β radiation for tumour ablation (Eβ, max = 1.84 MeV) and γ radiation for imaging (Eγ 80.6 keV, 6.71%) and has a half-life time of 26.8 h. In the in vitro, ex vivo and rabbit experiments, a batch of 18.7% weight by weight (Quiremspheres) was used, and in the human patients, a batch of 166Ho-MS with a Ho content of 17.6% weight by weight was used, corresponding to approximately 1 mg of Ho per 5.5 mg 166Ho-MS.
Phantom calibration for CT imaging
To quantify 165Ho-MS on CT imaging, series of HoCl3 and 165Ho-MS concentrations were made to create a calibration curve. A serial dilution (n = 10) of HoCl3 hexahydrate (HoCl3 • 6 H2O) (Sigma-Aldrich Chemie N.V., Zwijndrecht, The Netherlands) was made in sterile demineralised water (Versylene, Fresenius Kabi, B.V., Huis Ter Heide, The Netherlands) to obtain a homogeneous solution (range 0.0008–0.06 mmol/mL Ho). Therefore, a stock solution was prepared of 0.06 mmol/mL Ho, mixed for 5 min, and diluted by adding demineralised water. The concentration of the solutions was corrected for the mass percentage of Ho of the HoCl3 (43.47%).
Two batches of 1,200 mg 165HoMS, mass percentage 17.6% and 18.7% [14], were suspended in the injection solution containing 116 mmol phosphate (pH 7.2)-buffered saline with polyoxyethylene-polyoxypropylene block copolymer (Pluronic F-68, Sigma-Aldrich Chemie N.V., Zwijndrecht, The Netherlands) 2% weight per volume solution, and then diluted to ten concentrations (range 0.25–20.0 mg/mL165HoMS or 0.0016–0.12 mmol/mL Ho) of 165HoMS. The 165HoMS solutions were subsequently mixed 1:1 with a 2% agar (MP Agar, Roche Diagnostics, Almere, The Netherlands) solution to prevent settling due to the weight of these Ho microspheres (1.4 g/mL). Therefore, the agar powder was dissolved in sterile water (Versylene, Fresenius Kabi, B.V., Huis Ter Heide, The Netherlands) and heated to 90 °C for 10 min, resulting in a transparent fluid. Subsequently, a 5 mL Eppendorf tube was filled with 5 mL of the homogeneously distributed Ho microspheres in agar solution in a series (n = 10) with a concentration ranging from 0.125 to 10.0 mg/mL Ho. Once cooled to room temperature in approximately 5 min of continuous rotation to prevent settling, the agar became solid.
Ex vivo and in vivo administration of 166HoMS
Radioactive 166Ho-MS with a known specific activity (Bq/mg) were suspended in a solution of 2% weight by volume Pluronic® F-68 (Sigma-Aldrich Chemie N.V., Zwijndrecht, The Netherlands) in a 116-mmol phosphate (pH 7.2) buffer by gentle agitation and repeatedly drawing up and down in a 1 mL Luer-lock syringe (Becton Dickinson S.A., Madrid, Spain). The injections of the 166HoMS were performed with a 1-mL Luer-lock syringe through a 21G × 1½″ (0.8 × 40 mm) hypodermic needle (Becton Dickinson S.A., Madrid, Spain). The activity of the 166HoMS inside the syringes was measured using a calibrated dose calibrator (VDC-404; Veenstra Instruments, Joure, The Netherlands). The injection procedures were performed by the intratumoural Ho research team. Ex vivo and in vivo (laboratory animals) injections were performed by RCB, FN, and BvN. In human patients, injections were performed under ultrasound guidance by ML (nuclear physician) assisted by other members of the team.
Ex vivo CT quantification of 166Ho-MS
The feasibility of CT Ho quantification was evaluated in five samples of ex vivo chicken muscle tissue. Syringes with approximately 0.2 mL of 166Ho-MS suspension with increasing activity ranging from 15 to 81 MBq (corresponding to 3.4 to 18.4 mg Ho) were injected into the tissue samples. The actually injected amount of Ho (mg) was determined by the injected activity. To determine the injected activity, the syringes (before and after injection), the gauze with potential injection channel leakage, and the tissue sample were measured in the dose calibrator. Based on these measurements, the known specific activity of the 166Ho-MS and the weight and mass percentage of the 166Ho-MS as well as the injected amount (mg) of Ho was calculated.
Laboratory animals
All experiments were performed in agreement with “The Netherlands Experiments on Animals Act” (1977) and “The European Convention for the Protection of Vertebrate Animals used for Experimental Purposes” (Strasbourg, 18.III.1986). Approval was obtained from the Utrecht University Animal Experiments Committee (DEC 2011.III.08.080).
The VX-2 tumour model in New Zealand white rabbits was previously described [15]. In short, in one animal, the donor rabbit, a tumour was implanted by injection of a suspension of approximately 4.0 × 107 VX-2 carcinoma cells subcutaneously into both flanks. Single tumours were induced in each of five rabbits by harvesting the tumour from the donor rabbit. A subcutaneous injection of 3 ± 1 mm3 viable fragments of VX-2 carcinoma with 0.1–0.3 mL phosphate-buffered saline was performed into the flank of five adult female New Zealand White rabbits weighing 3–4 kg. All tumour implantations and treatments were performed under analgesia with carprofen 4 mg/kg. During the animal experiments, sedation and analgesia were achieved with a mixture of 0.125 mg/kg dexdomitor and 15 mg/kg ketamine.
After the intratumoural injection of 0.2 mL of 166Ho-MS suspensions in five rabbits, a CT scan was performed according to the clinical protocol described below. Only in rabbit number 5, a higher amount of activity (57.9 MBq) was injected for comparison to quantitative SPECT imaging and depositions outside the tumour. CT and SPECT data were compared to the injected mg Ho, which was determined by the injected activity as described above.
Patients
Human patients previously treated with direct intratumoural injections of 166Ho-MS were analysed to provide an example of clinical CT quantification method. Between 2015 and 2017, four patients with head and neck cancer were referred by their head and neck oncologist. Three patients were treated in a palliative setting. If no other palliative treatment options were available, and nonetheless a strong wish for treatment existed, patients were amenable for direct intratumoural injections of 166Ho-MS, with the aim of improving the patients’ quality of life. One patient was part of a prospective clinical pilot study (NCT02975739) [8]. All patients provided informed consent before treatment. Immediately after two to four ultrasound-guided intratumoural injections, containing a total of 100 mg of 166Ho-MS, a SPECT, a high-dose CT, and a planar scintigraphy of thorax and abdomen (imaging time 300) were performed. This retrospective analysis of patients treated in a palliative setting was approved by the medical ethical committee of the University Medical Center Utrecht. Some study subjects or cohorts have been previously reported [8].
CT and SPECT
All CT and SPECT imaging was performed using a Symbia T16 SPECT/CT system (Siemens, Erlangen, Germany) that combines a dual-headed gamma camera with a 16-slice CT system. The acquisition parameters were identical to a diagnostic high-dose CT scan of the head and neck region in the clinical setting: tube voltage 110 kVp, effective tube current 225 mAs; detector configuration 16 × 0.6 mm; rotation time 0.6 s; helical scan mode; pitch 1.0. Images were reconstructed with a 1.5-mm slice thickness with a 0.7 mm increment (voxels size 0.56 × 0.56 × 0.7 mm) and a B31s medium smooth reconstruction kernel.
Medium-energy low-penetration collimators were used on both SPECT cameras. Energy windows were set at 80.6 keV (15% window width) for the 166Ho photopeak. A total of 120 projections of 30 s were acquired in a 360° noncircular orbit. Quantitative image data were reconstructed to a 1283 matrix with an isotropic voxel size of 4.8 mm3. The reconstructions were performed using previously validated Monte-Carlo-based reconstruction software [2, 16] using an ordered subsets expectation maximisation algorithm (10 iterations with 8 subsets) and a quantitatively correct forward model, resulting in an absolute quantitative three-dimensional activity distribution in MBq/voxel [2]. Based on the recovered activity and the specific activity of the microspheres, the absolute amount of Ho (mg) was calculated.
Data analysis
CT data were analysed with ImageJ version 1.50b (NIH, Bethesda, USA). In the phantom of both HoCl3 and Ho-MS, a circular region of interest (ROI) with a 10-mm diameter was drawn in the centre of the 5-mL Eppendorf tube. This ROI was applied over 30 slices to create a volume of interest of 1.6 cm3. Subsequently, HU value of this VOI (mean ± standard deviation [SD]) was calculated. A scatterplot of the observed HU values against the calculated concentration values was made to obtain a calibration curve.
In the chicken muscle tissue experiments, an VOI was drawn in an area without microspheres. From this VOI, the mean ± SD and maximum HU values were obtained. In the animals and patients, the tumour volume was manually segmented on CT images by one of the investigators (RCB). Based on the literature, it was assumed that all tumour voxels had a HUvalue between -50 and 100 [17] and that all voxels with a HU > 100 contained Ho microspheres. The amount of Ho was calculated using the following strategy:
- 1.
All voxels with HU > 100 were selected and divided by the HU/concentration calibration curve, to obtain a concentration of mg/ml/voxel;
- 2.
This concentration was multiplied by the voxel volume (0.22 mm3) to calculate the absolute amount of Ho per voxel;
- 3.
The total sum of these voxels resulted in the total amount of Ho in milligrams.
Continuous data were presented as mean ± SD deviation if normally distributed and as the median and range if skewed (Shapiro-Wilk test). A Pearson correlation coefficient between the dose calibrator, CT-, and SPECT-quantification was calculated if normally distributed, with Spearman rank-order correlation if skewed. Agreement between the measurements was presented as the recovered percentage on SPECT or CT compared to the dose calibrator or SPECT. SPSS software (SPSS for Windows, version 22.0; SPSS Inc., Armonk, USA) was used for all analysis.