The present study was designed to establish the 18F-AzaFol PET/CT dosimetry as a radiation safety cohort in patients with adenocarcinoma of the lung. It is part of a larger, prospective, multi-centric study conducted in three Swiss hospitals. The study was approved by the respective institutions’ ethics committee, the national health authorities, centrally monitored by a dedicated, independent Contract Research Organization and listed in the trial list of the National Institute of Health Trial database (NCT03242993).
Eligible patients had a histologically confirmed non-small cell lung carcinoma of the lung with measurable lesions (≥ 10 mm according to RECIST 1.1 ) and were staged by standard of care imaging with an indication for systemic treatment. The last systemic treatment should not have been applied within 3 weeks before performing the study exam. Male and female patients needed to be 18 years or older as well as voluntarily sign the informed consent form. Exclusion criteria were, in brief, any contraindications to the class of drugs under study, pregnancy or breast-feeding, clinically significant concomitant disease states (e.g., renal failure, hepatic dysfunction, or cardiovascular disease) and poor performance status (ECOG > 2). Patients were asked to discontinue taking vitamin supplements containing folic acid 48 h prior to the PET/CT. Due to pre-clinical results established by our group, all patients received a single bolus injection of 1 mg of folic acid before tracer administration .
The precursor (protected 3′-aza-2′-chloro-folate) was synthesized by an external manufacturer (Merck & Cie, Schaffhausen, Switzerland). Radiolabeling and subsequent global deprotection of the protected 3′-aza-2′-chloro-folate precursor resulted in 3′-aza-2′-[18F]fluorofolic acid herein referred to as 18F-AzaFol. Radiolabeling of the folate precursor was achieved within 17 min at 150 °C in dimethyl sulfoxide (≥ 99.7%, over molecular sieve, Acros Organics) and the protective groups were subsequently cleaved under acidic conditions (4 M hydrochloric acid, Merck EMPROVE®) at 60 °C. The product was purified over two solid-phase extraction cartridges (30 mg mixed-mode cation exchange sorbent for bases, particle size 30 μm (MCX), Oasis) and subsequently eluted with 50 mM phosphate buffer containing 10% methanol (pH 7.4). The desired compound was then isolated by isocratic HPLC (eluent, 20 mM phosphate buffer containing 7% methanol pH 7.4; column, Luna 5 μm PFP(2) 100 Å, 250 × 10 mm, Phenomenex; flow rate, 4 mL/min.; retention time, 11 to 12 min). After further purification over a third MCX cartridge, formulation was achieved by eluting the product from the MCX cartridge with 5 mL of 50 mM phosphate buffer containing 10% ethanol (pH 7.4) into the bulk vial, which was prefilled with 9 mL of saline. The bulk product was then aseptically dispensed through a sterile filter (Millex-GV, 0.22 μm, PVDF, diameter 33 mm, Millipore) to yield the final, sterile, and non-pyrogenic product with a total activity of approximately 2 GBq in 6 mL plus samples for quality control, including sterility testing and a retained sample. Radiochemical purity was ≥ 95% and molar activity ≥ 100 GBq/μmol. The analytical HPLC was run with an isocratic method (column, XSelect HSS PFP XP, 100 Å, 2.5 μm, 4.6 mm × 150 mm (Waters); eluent, 10 mM ammonium acetate buffer, pH 6.2 containing 5% methanol; flow rate, 1 mL/min; UV, 280 nm; radio detector, 2 × 2″ NaI (Tl) crystal). The retention time of 18F-AzaFol was 7.8 min.
PET/CT acquisition protocol
Seven PET images (from the top of the skull to the mid femora, 1 min/bed position, six bed positions, total scan length was 6 min) were acquired on a Discovery MI time-of-flight (TOF) PET/CT (GE Healthcare, Waukesha, WI, USA) at the University Hospital Zurich. All pertinent corrections (normalization, dead-time, physical decay correction at the start of the PET scan, random coincidence, attenuation, and scatter correction) were applied. PET field-of-view width was 70 cm with 256 × 256 image discretization. The voxel size in the axial direction and transverse plane were, respectively, 2.79 mm and 2.73 mm. Each patient underwent a low-dose (120 kVp, 15 mA/s, spiral pitch = 0.98, 40 mm beam collimation, CTDIvol = 1.3 mGy) whole-body CT for attenuation correction prior to the first PET acquisition.
PET data were reconstructed using a proprietary three-dimensional ordered subset expectation-maximization algorithm (3 iterations × 16 subsets) with TOF information and PSF recovery correction (VPFXs, vendor-based reconstruction algorithm) with 6.4-FWHM Gaussian post-reconstruction filtering.
Patients were instructed to void the urinary bladder after the 1 h scan period.
Co-registered PET and CT data were loaded using PMOD 3.9 (PMOD Technologies, Zurich, Switzerland). Volumes of interest (VOI) were manually drawn slice-by-slice on the axial plane of the CT part of each PET/CT study using the polygonal segmentation tool of PMOD by two operators in consensus (SG, NS) for the following body regions: brain, thyroid, lungs, heart, liver, spleen, stomach, kidneys, prostate (in men), red marrow, intestines, and whole-body. Tumors with increased folate receptor expression were manually segmented on CT data using the combined information of PET/CT and prior standard of care clinical imaging.
The urinary bladder was manually segmented on the emission PET data to account for possible changes of volume due to bladder filling between PET/CT acquisitions.
Specific biological uptake was found in the choroid plexuses. This small vascular structure was segmented by emission-based threshold segmentation to avoid important PET signal spill-out. We adopted a threshold of 5% of the maximum signal intensity to delineate the choroids VOI. This relatively low threshold level was possible considering the negligible tracer uptake of the surrounding cerebral tissue. An example of source organ VOI segmentation is provided in Supplemental Data 1 (Figure S1).
Absorbed dose (AD) estimations
The total activity contained in each considered source organ was obtained by multiplying the average activity concentration (expressed in Bq/mL) by the organ volume expressed in milliliters and normalized to the administered patient activity at each time point. For all source organs, normalized time-activity curves (nTAC) were obtained assuming an initial constant uptake (Aorgan (t) = Aorgan (t = 3 min) for 0 min ≤ t < 3 min) and thus using a bi-exponential fit of experimental data extended from t = 3 min and the last measured data point (60 min post-administration). Exponential fit parameters were obtained using the kinetic module of OLINDA/EXM v.2.1. Beyond, in absence of late measured data, a simple mono-exponential physical decay was assumed to derive time-integrated activity coefficients (TIACs) by analytical time integration of source organ time-activity curves using MATLAB software (release 2017a; The MathWorks, Inc., Natick, MA, USA).
Bone marrow dosimetry was estimated by combining information from different VOIs as previously reported . In brief, the red marrow time-integrated activity coefficient (TIAC) was obtained from nTAC in which the red marrow activity concentration (Bq/mL) was sampled in VOIs drawn in the humeral bone, the heads of femora, and lumbar vertebrae (L3–L4). The total activity in the red bone marrow was obtained by multiplying the average activity concentration measured in the sampled VOIs by the total red marrow mass of ICRP-89 adult male/female reference phantoms , which was repeated for each patient and at each acquired time point.
To estimate the colon AD, the total number of colonic disintegrations was partitioned to its components (right colon, left colon, and rectum) proportionally to their respective masses of the ICRP-89 male and female reference phantoms .
The total amount of radioactivity excreted from the body through the urine was not quantitatively assessed as no urine samples were collected in this study. We estimated the total amount of radioactivity present in the urinary bladder at 1 h after administration (latest quantitative PET acquisition available). This amount of radioactivity was assumed to be completely voided just after the last PET acquisition, when the patient was allowed voiding at the toilet. To obtain the TIAC for the urinary bladder, the excreted fraction (the voided activity at 1 h divided by the patient-administered activity) was used in input to the special kinetic module of OLINDA/EXM 2.1 for the urinary bladder voiding using a voiding period of 1 h. We adopted this methodology in the absence of urine samples and PET acquisition for t > 60 min.
The TIAC for the rest of the body was obtained by subtracting the sum of the source organ TIACs from the whole-body TIAC.
TIACs were used in input to the OLINDA/EXM® 2.0 code (HERMES Medical Solution AB, Stockholm, Sweden)  that provided organ AD and effective dose (E) per unit of administered activity in μGy/MBq and μSv/MBq, respectively, using the NURBS voxel-based phantoms  adjusted to the ICRP-89 organ masses  and ICRP103 tissue weighting factors (wT) . According to publication ICRP 103, E is calculated as the weighted average of organ/tissue equivalent doses, summing equivalent doses multiplied by tissue weighting factors (wT), which provide a simplified representation of fractional contributions to total stochastic detriment from cancer and heritable effects. While exposures may relate to individuals or population groups, E is calculated for reference persons. For a general population, ICRP recommends to average E computed for a reference male and a reference female phantom. We will call this quantity “gender-averaged E.” In this paper, we will also compute E for specific patients, which we will refer as “patient E” using the reference organ masses of OLINDA/EXM 2.0, thus adopting a methodological approach typical of radiation protection, where the dosimetry of a reference adult subject is the focus.
A specific dose assessment was performed for the choroid plexuses and tumors, which do not appear in the list of available source/target organs in OLINDA. Trapezoidal time integration of the normalized time-activity curve, for 0 ≤ t ≤ 60 min, was performed using MATLAB software (release 2017a; The MathWorks, Inc., Natick, MA, USA), analytical integration assuming mono-exponential physical decay was applied beyond. To account for loss of signal due to partial volume effects (PVE), tumor TIACs were corrected using recovery coefficients (RC) obtained in a NEMA/IEC NU2 phantom experiment using the same clinical acquisition and reconstruction parameters used for the patient study as reported in Supplemental material (Supplemental Data 2). TIACs for the choroid and tumors were used as input in the OLINDA sphere model. For the choroids, we also applied a more realistic AD estimation using the Monte Carlo-derived analytical approach proposed by Amato et al.  in which the complex geometry of this organ was approximated using a set of simple geometrical structures such as (cylinders and parallelepipeds). The mass of the choroids was estimated to be 2.76 g and 1.81 g for male and female as previously reported in .
We performed an additional analysis to investigate the influence of the nTAC extrapolation to infinity on source organ TIACs. At this scope, in addition to the assumption of pure physical decay from t > 60 min, we also computed mean source organ TIACs (obtained from the time integration of averaged nTACs across the six patients) by assuming mono- and bi-exponential prolongation to infinity of the nTACs.
Coefficients of determination (R2) were computed to evaluate the goodness of both mono-exponential and bi-exponential nTAC fits for source organs exhibiting a monotonic decreasing uptake or decreasing activity accumulation with time (all considered source organs except the urinary bladder).
We characterized the tumor to background contrast as a function of the time elapsed after radiotracer administration; with this aim, we computed for each acquired PET the tumor to lung ratio (T/L) by dividing the average activity concentration measured in lung tumor VOIs by the average activity concentration measured in the lung VOI. We report the average ± SD of T/L obtained for nine tumors.