Patient population and imaging data
Patients diagnosed with GEP-NET who received a whole-body [68Ga]Ga-DOTATATE or [68Ga]Ga-HA-DOTATATE PET/CT between August 2011 and April 2016 were included. Patients without apparent NET-lesions on diagnostic PET/CT were allocated to a separate cohort that was used as a control group, which represented patients that were referred with suspected NET or received treatment for prior NET. Informed consent was obtained via institutional procedures from all individual participants included in the study.
Patients were treated according to the EANM-guidelines [
29], and SSA therapy was withdrawn prior to [
68Ga]Ga-(HA-)DOTATATE administration (where (HA-)DOTATATE refers to both peptides).
68Ga-labeled (HA-)DOTATATE was produced in-house according to validated procedures described previously [
30]. An intravenous injection of approximately 100 MBq
68Ga-labeled (HA-)DOTATATE was administered, and at ~ 45 min post-injection a scan was acquired on a Gemini TF PET/CT (Philips, the Netherlands) with 2–2.5 min per bed position from skull to thighs (axial field of view of 18 cm). Image data was reconstructed using the Philips specific BLOB OS-ToF algorithm with ‘normal’ smoothing (isotropic 4 mm voxels).
SUVpeak, SUVmax and radioactivity concentrations (MBq/L) for aorta, spleen, thyroid, liver, primary tumor and metastatic lesions > 2 cm in diameter were determined by placing spherical volume of interests (VOIs) of at least 2 cm. Then, all radioactivity concentrations were corrected for decay to time of injection and subsequent peptide concentrations (µg/L) per organ or tumor were calculated based on the measured radioactivity concentrations and the administered specific activities (MBq/µg).
To assess the effect of tumor burden on biodistribution, patients were divided into three groups based on quantity and location of metastases on their clinical PET/CT report. These three groups were: (1) patients with primary tumors or with ≤ 2 metastases (limited disease), (2) patients with liver-only metastases, and (3) patients with extensive metastases (i.e. , not classified as group 1 or 2). The PBPK model predictions, for both [68Ga]Ga-DOTATATE and [68Ga]Ga-HA-DOTATATE, were evaluated with clinical imaging data for these three groups separately.
PBPK model development
A PBPK model for GEP-NET patients was developed in PK-Sim
® and MoBi
® (Open Systems Pharmacology software, version 9.0) using the standardized protein base model [
31]. Compound-specific information includes the physicochemical parameters of the compound, and this was separately added for DOTATATE and HA-DOTATATE to obtain compound-specific predictions. Eventually, predictions were performed separately for each patient group, based on metastatic status, for both peptides.
All input parameters for peptide specific distribution and uptake, including intracellular degradation rates and SSTR2 amounts, were based on the previous described [
68Ga]Ga-DOTATATE PBPK model [
26]. Compound-specific parameters for DOTATATE were also derived from the previous published PBPK model [
26], while compound-specific input parameters of HA-DOTATATE (molecular weight and lipophilicity) were based on literature. The input parameters molecular weight and lipophilicity for [
68Ga]Ga-HA-DOTATATE were 1628.5 g/mol and −3.12, respectively [
32]. Plasma protein binding for [
68Ga]Ga-HA-DOTATATE was unknown, but a major difference compared to [
68Ga]Ga-DOTATATE was not expected; thus, the fraction unbound was set at 0.69, accordingly. Renal clearance was manually scaled to a predicted 13% unchanged excretion in urine within the first 2 h post-injection for all patients [
33]. For the predictions per group for both peptides, the mean estimated glomerular filtration rate (GFR) per group was used as renal function input (calculated using the Cockcroft–Gault equation [
34]). In addition, all other assessed patient characteristics (i.e., age, weight and height) were used as system-specific input parameters. Internalization rates were assumed to be similar for organ and tumor tissue and were fixed to 0.268 min
−1. This was based on a 1.67-fold increase in internalization rate for [
68Ga]Ga-HA-DOTATATE compared to [
68Ga]Ga-DOTATATE (accumulation of [
68Ga]Ga-HA-DOTATATE in spleen plateaued 30 min post-injection compared to 50 min post-injection for [
68Ga]Ga-DOTATATE) [
26,
33,
35]. Unknown input parameters for HA-DOTATATE (the equilibrium dissociation constant (
KD) and dissociation rate constant (
koff)) were optimized by using all observed spleen concentrations from all included patients at once and thus optimized values represent
KD and
koff for the entire population. By using observed data of spleen, which expresses SSTR2 but not SSTR5 [
32,
36], it was ensured that the optimized affinity values of HA-DOTATATE were specific for SSTR2. Model fitting was performed using a built-in Monte Carlo algorithm for parameter identification to optimize selected input parameter to describe the data best.
Separate tumor compartments were added to the model to describe distribution to these compartments and the effect of this tumor uptake on normal organ uptake. Since physiological tumor characteristics can differ between primary tumors and its metastases, three compartments were added: primary tumor, liver metastases and other metastases. All metastases other than liver metastases were gathered in one compartment since only small uptake differences were observed in clinical practice between those metastases and also for reasons of model simplicity. Tumor physiology characteristics for all three tumor compartments were based on literature. The tumor volume for the primary tumor was set at 10 mL for all groups, and liver metastases were assumed to have a total volume of 50 mL. Volumes of other tumor metastases differed for each patient group and were set at 5 mL, 0 mL and 50 mL for groups 1, 2 and 3, respectively, based on clinical observations indicating a mean total tumor volume of approximately 65 mL in 232 patients [
37]. Although quantification of tumor volumes in this article was debatable, since there is no gold standard for tumor volume measurements, yet a mean total NET volume of approximately 65 mL reflected the population of patients treated in our hospital. Besides, this was comparable to median tumor volumes that were used (or fitted) in previous PBPK models [
21,
27,
38,
39]. Vascular sub-compartments within tumors were estimated based on literature values for primary tumors and liver metastases, resulting in fraction vascular of 0.21 and 0.17, respectively [
40‐
45]. Fraction vascular for other metastases was unknown and fixed to 0.075 based on previous PBPK models [
21,
22,
46]. The interstitial fraction was assumed to be similar for each tumor and was fixed to 0.3 [
22]. Blood flow was set at 152 mL/min/100 g for the primary tumor compartment [
45]. In absence of relevant data for distant metastatic sites, the same value was used for these sites. For liver metastases, a higher blood flow of 203 mL/min/100 g was used as an input parameter, due to their nature of hyperperfusion [
42‐
44]. SSTRs were added to the tumor compartments with an expression relative to spleen. Since NET metastases showed higher SSTR2 expressions compared to primary tumor sites [
47], a higher relative expression was added to this compartment (relative value of 1.5 for primary tumors and 2 for all metastases). These fractional differences between tumors compared to spleen were in accordance with previous NET PBPK models [
21,
22]. Transcapillary exchange of the radiopharmaceutical from plasma to the interstitial space was described by the two-pore formalism [
46]. Pore radii for small and large pores were fixed to 4.5 and 500 nm, respectively, representing a leaky tumor vessel capillary [
46,
48]. The flow fraction via the large pores was 0.8, reflecting a discontinuous membrane as expected in tumors [
46]. Hydraulic conductivity describes passage of porous material and was fixed to a literature value of 0.00126 mL/N/min for tumors [
49,
50].
Minimum and maximum ranges for administered peptide amounts per group and SSTR2 organ expression (derived from the previously published PBPK model) were imputed to derive prediction intervals for organ and tumor radionuclide distributions [
26]. In addition, to take into account a high variability in tumor uptake, prediction intervals in all three tumor compartments were based on observed (inter)quartile ranges in blood flow and blood volumes of the specific compartment [
42‐
45]. For other metastases, blood flow was assumed to be comparable to primary tumors and fraction vascular was assumed to vary from 50 to 150% relative to the median. This resulted in minimum and maximum blood flow of 53 and 252 mL/min/100 g for primary tumors and other metastases and 139 and 363 mL/min/100 g for liver metastases. Fraction vascular ranged from 0.09 to 0.34 for primary tumors, 0.1 to 0.23 for liver metastases and 0.025 to 0.125 for other metastases.
A sensitivity analysis was performed to calculate the sensitivity of the PK model output for certain parameter assumptions using a build-in algorithm, which was performed by alteration of input parameters with ± 10% [
51]. This sensitivity (
Si,j) is calculated using the following equation:
$$S_{i,j} = \frac{{\Delta {\text{PK}}_{j} }}{{\Delta p_{i} }}*\frac{{p_{j} }}{{{\text{PK}}_{j} }}$$
(1)
where
PKj is a PK parameter of a certain output to an input parameter (
pj)
. Thus, the sensitivity for the PK parameter to that input parameter was calculated as the ratio of the relative change of that PK parameter (
ΔPKj) and the relative variation of the input parameter (
Δpi). A sensitivity value of − 1 implies that a 10% increase of the input parameter resulted in a 10% decrease of the PK parameter output.
DOTATATE versus HA-DOTATATE
Administered patient doses of [68Ga]Ga-DOTATATE and [68Ga]Ga-HA-DOTATATE are based on radioactivity (~ 100 MBq per administration). Specific activities (MBq/µg) were derived from the tracer production logs to calculate the administered absolute peptide amount. However, specific activities, and thus administered peptide amounts, differed between production batches and peptides. In general, this resulted in higher administered total peptide amounts for DOTATATE compared to HA-DOTATATE.
It is of interest to directly compare organ uptake between both peptides to assess (dis)similarities in this organ accumulation, without discrepancies in administered peptide amounts that could alter such a comparison. Due to a change in production procedures in 2017, an additional group of GEP-NET patients receiving higher peptide amounts of [68Ga]Ga-HA-DOTATATE (comparable to administered DOTATATE amounts) could be selected for a subanalysis to exclude a potential effect of different administered peptide amounts for [68Ga]Ga-DOTATATE and [68Ga]Ga-HA-DOTATATE on organ uptake. The sample size of this additional group was matched to the included patients for PBPK evaluation receiving HA-DOTATATE, and spleen was used as a reference organ to compare uptake (SUVpeak and SUVmax).