Intra-individual comparison of 68Ga-PSMA-11 and 18F-DCFPyL normal-organ biodistribution

Purpose Detailed data comparing the biodistribution of PSMA radioligands is still scarce, raising concerns regarding the comparability of different compounds. We investigated differences in normal-organ biodistribution and uptake variability between the two most commonly PSMA tracers in clinical use, 68Ga-PSMA-11 and 18F-DCFPyL. Methods This retrospective analysis included 34 patients with low tumor burden referred for PET/CT imaging with 68Ga-PSMA-11 and subsequently 18F-DCFPyL. Images were acquired with 4 cross-calibrated PET/CT systems. Volumes of interest were placed on major salivary and lacrimal glands, liver, spleen, duodenum, kidneys, bladder, blood-pool and muscle. Normal-organ biodistribution of both tracers was then quantified as SUVpeak and compared using paired tests, linear regression and Bland-Altman analysis. Between-patient variability was also assessed. Clinical and protocol variables were investigated for possible interference. Results For both tracers the highest uptake was found in the kidneys and bladder and low background activity was noted across all scans. In the quantitative analysis there was significantly higher uptake of 68Ga-PSMA-11 in the kidneys, spleen and major salivary glands (p <  0.001), while the liver exhibited slightly higher 18F-DCFPyL uptake (p = 0.001, mean bias 0.79 ± 1.30). The lowest solid-organ uptake variability was found in the liver (COV 21.9% for 68Ga-PSMA-11, 22.5% for 18F-DCFPyL). There was a weak correlation between 18F-DCFPyL uptake time and liver SUVpeak (r = 0.488, p = 0.003) and, accordingly, patients scanned at later time-points had a larger mean bias between the two tracers’ liver uptake values (0.05 vs 1.46, p = 0.001). Conclusion Normal tissue biodistribution patterns of 68Ga-PSMA-11 and 18F-DCFPyL were similar, despite subtle differences in quantitative values. Liver uptake showed an acceptable intra-patient agreement and low inter-patient variability between the two tracers, allowing its use as a reference organ for thresholding scans in the qualitative comparison of PSMA expression using these different tracers. Electronic supplementary material The online version of this article (10.1186/s40644-019-0211-y) contains supplementary material, which is available to authorized users.


Introduction
Prostate-specific membrane antigen (PSMA) is a transmembrane glycoprotein upregulated in most prostate cancer (PC) cells and its expression levels relate to tumor stage and grade [1][2][3]. Exploiting this feature, several low-molecular-weight Glu-ureido-based PSMA inhibitor radioligands for Positron Emission Tomography/Computed Tomography (PET/CT) have entered the clinical scope of PC workup [4][5][6]. These tracers specifically bind to the extracellular domain of PSMA and are internalized, leading to tumor uptake and retention [7,8].
The most widely used PSMA radiotracer is 68 Ga-PSMA-11 (also named 68 Ga-PSMA-HBED-CC) [9], with increasing clinical experience in a variety of PC indications [10][11][12][13][14]. Despite the promise and widespread clinical adoption of this agent, there are logistic issues with use of this tracer related to its short physical half-life (68 min) and decreasing synthesis yields as generators decay. It is also difficult and expensive to comply with good manufacturing practice guidelines and therefore centralized radiopharmacy production and distribution are constrained. These issues have encouraged development of 18 F-labeled PSMA ligands [15][16][17], which allow large-scale, cyclotron production and distribution to meet growing demand for molecular imaging evaluation of PC. Among these, the most common agent in clinical use is 18 F-DCFPyL, which is a second-generation fluorinated PSMA-targeted PET radiotracer [15] already showing promise clinically [18][19][20][21][22].
Besides qualitative assessment of the presence of malignant PSMA-expressing lesions, PSMA PET/CT has an evolving role in the quantitative evaluation of tumor target expression, with potential applications in prognostic stratification, assessment of suitability for PSMA-targeted therapy and subsequent evaluation of treatment response [23][24][25][26][27]. With expected increasing adoption of PSMA-ligand PET/ CT into clinical practice and trials, reporting standards are being developed to enhance reproducibility and communication with clinicians [28]. For this purpose, PSMA expression categories could be defined in relation to reference organs including blood pool, liver and parotid glands. Reference organs have also been used for assessment of suitability of patients for PSMA-targeted therapy [29]. However, detailed data on comparison of biodistribution of different PSMA-ligands is scarce and molecular diversity across the range of available radiotracers is expected to impact tissue kinetics [5,30], raising concern regarding the generalization of observations and comparability of quantitative data.
In this study we explored 68 Ga-PSMA and 18 F-DCFPyL biodistribution in a routine clinical setting, making an intra-individual comparison of normal-organ uptake between the two radiotracers. Patients with low tumor burden were selected to minimize the impact of tumor-sink effect on normal tissue biodistribution.

Patients
We retrospectively reviewed 43 consecutive adult male patients who had undergone PET/CT scans with both 68 Ga-PSMA-11 and 18 F-DCFPyL at Peter MacCallum Cancer Centre, between October 2014 and April 2018. All scans were performed as part of routine clinical work-up. From this cohort, 34 patients were included in this study according to the following criteria: 1) low tumor burden, defined as a prostate-specific antigen (PSA) < 20 ng/mL; 2) stable disease between scans, defined as an interval PSA change < 10 ng/mL; 3) no appreciable altered biodistribution on imaging, taking into account absence of radiopharmaceutical infiltration, sink effect due to high tumor burden or renal impairment. We also recorded clinical data, namely baseline PC characteristics, PSA level at the time of each scan and previous therapies. This research has been approved by the institutional ethics committee and patient consent was waived (approval number: 15/46R).

Image analysis
Image analysis was performed by a Nuclear Medicine physician using an appropriate workstation and vendor neutral software (MIM Encore™ version 6.7, MIM Software Inc., Cleveland, USA). Volumes of interest (VOIs) were automatically drawn over entire organs of moderate to intense physiologic uptake and/or tracer accumulation, namely the major salivary and lacrimal glands, duodenum (third portion), spleen, kidneys (cortex) and urinary bladder, using a gradient-based contouring tool (PET Edge®). Additionally, spherical VOIs were drawn inside the parenchyma of the right hepatic lobe (6 cm diameter), descending thoracic aorta and right gluteus muscle (each 2 cm diameter) - Fig. 1. Tracer biodistribution was then quantified by the peak standardized uptake value (SUV peak ) as defined in PERCIST criteria [34], which yields less intra-patient bias compared to SUV max and, unlike SUV mean , does not require definition of tumor boundaries [35]. For paired organs, the arithmetic mean is presented.

Statistical analysis
Statistical analysis was performed using SPSS 25.0 (IBM Corp., Armonk, NY) and GraphPad Prism 7 (GraphPad Software, La Jolla, USA). Normality was tested using the Shapiro-Wilk test. Intra-patient comparison of quantitative uptake between the two tracers in each of the target organs encompassed Paired t or Wilcoxon signed-rank tests, linear regression and Bland-Altman analysis. Inter-patient variability was assessed by taking the coefficient of variation (COV). Interference of clinical and protocol variables was also investigated, using Spearman's rank correlation, and by subgroup analysis comparing the quantification biases in a group of patients where imaging with the two tracers was done within a similar time-frame and another with higher 18 F-DCFPyL uptake duration (Independent samples t or Mann-Whitney U test). To account for multiple tests, the significance level used was p ≤ 0.01 (two-tailed).

Results
Patient characteristics are summarized in Table 1. Visual assessment of the expected biodistribution of both tracers was considered normal for all but one patient, who had a past history of right nephrectomy prior to both scans. There was no evidence of malignant involvement within the target organs. Most patients (67.6%) had low volume malignant PSMA expressing lesions on either scan (prostatic/prostate bed in 20.6%, nodal in 44.1% and bony in 17.6% of patients). The remaining 11 (32.4%) patients had no evidence of disease on both exams. The median absolute difference in PSA levels at the time of each scan was 2.75 (IQR 0.43-3.18) ng/mL.

Organ uptake comparison
Normal-organ biodistribution was grossly equivalent for both tracers. The highest activities were observed in the kidneys and bladder, followed by the salivary glands. Liver, spleen and proximal small bowel also showed prominent uptake using both tracers and low background activity was noted in the blood-pool (thoracic aorta) and muscle across all scans. These observations conformed with the quantitative uptake values (SUV peak ) for each tracer (Fig. 2).
Despite the visually appreciable similarities, there were subtle but statistically significant differences in the quantitative analysis of normal-organ uptake amidst the two agents. Detailed comparison is described in Table 2 and depicted in Figs. 3 and 4. Liver tracer quantification between tracers was well correlated, allowing for some variability on a per-patient basis (Fig. 3d). Quantitative uptake was slightly higher in 18 F-DCFPyL scans (mean SUV peak 7.5 vs 6.7, p = 0.001), associating with a mean bias of 0.79 ± 1.30 (Fig. 4d). In contrast, the spleen presented significantly higher 68 Ga-PSMA-11 uptake values (median SUV peak 9.4 vs 4.9, p < 0.001). While splenic activity quantification also correlated well (Fig. 3e), there was an increasing proportional bias at higher uptake values (Fig. 4e, r = 0.681, p < 0.001, mean bias − 4.49 ± 1.78). Tracer activity in the renal cortex was also significantly higher in 68 Ga-PSMA-11 scans (mean SUV peak 59.6 vs 40.0, p < 0.001), with a mean bias of − 19.60 ± 9.52 (Fig. 4g). Lacrimal and major salivary glands had a good correlation in quantitative uptake between the two scans ( Fig. 3a, b, c) and an acceptable overall agreement with a calculated mean bias of − 0.39 ± 1.49 in lacrimal, − 2.08 ± 2.39 in parotid, and − 3.21 ± 2.46 in submandibular glands (Fig. 4a, b, c).
Within the solid organs, the lowest uptake variability between patients was found in the liver (COV 21.9% for 68 Ga-PSMA-11 and 22.5% for 18 F-DCFPyL).

Variables influencing organ uptake
Age, weight, tracer dose (MBq/Kg), uptake time and PSA level were tested for possible correlation with each tracer's normal-organ uptake (Additional file 1: Table S1). In 18 F-DCFPyL scans, there was a weak correlation (r = 0.488, p = 0.003) between uptake time and liver uptake values (Fig. 5) and lacrimal glands SUV peak (r = 0.554, p = 0.001). None of the tested variables correlated with 68 Ga-PSMA-11 uptake within the target organs.
Agreement of uptake quantification between the two tracers was further investigated, taking into account 18 F-DCFPyL uptake time variability by conducting subgroup analysis (Additional file 1: Table S2). The group of patients scanned with both tracers within a similar time-frame (up to 90 min after injection, n = 16) had a statistically significant lower mean bias between the two tracers in liver uptake quantification (0.05 ± 1.122 vs 1.46 ± 1.093, p = 0.001). Furthermore, in both these groups liver COVs of 18 F-DCFPyL scans were distinctly lower compared to those of the entire study population (16.5 and 18.4% vs. 22.5%), while in 68 Ga-PSMA-11 scans liver uptake COVs were still similar (20.4 and 22.6% vs. 21.9%).

Discussion
In this intra-individual comparison of patients scanned with 68 Ga-PSMA and 18 F-DCFPyL, the overall biodistribution in normal organs was similar, with both tracers showing specific retention in the salivary and lacrimal glands, small intestine, liver, spleen, kidneys and bladder. While this was somewhat expected, subtle but statistically significant differences were found regarding normal-organ uptake. The highest uptake was observed in the urinary system in keeping with a predominantly renal clearance for both tracers [9,19,36]. Kidney tracer retention was, however, higher in 68 Ga-PSMA-11 scans, which could relate to specific cortical binding and slower renal clearance for this compound [37]. On the other hand, bladder SUV was higher in 18 F-DCFPyL scans. This comparison is particularly relevant, as local relapses are common and a diagnostic challenge in the work-up of biochemical recurrence. While slower clearance of 68 Ga-PSMA-11 could be a possible explanation for this finding, variable hydration, voiding status and non-uniform use of iodinated contrast are important confounders that render interpretation difficult, which is reflected in the very high bladder COV for both scans. Despite this potential limitation, the longer physical half-life of 18 F allows for delayed imaging with the opportunity for dilution of urinary activity and delayed post-void imaging.
Salivary and lacrimal glands also showed intense uptake and good correlation between the tracers, which increases confidence in inter-scan comparability. Additionally, it was noted that lacrimal gland uptake, unlike salivary gland uptake, seemed to be dependent on 18 F-DCFPyL uptake time. This is congruent with the hypothesis that later time-point acquisitions may increase the detectability of small structures, which has recently been demonstrated for PC lesions   [17,39]. Blood-pool and muscle background activity was very low, providing excellent image quality with both tracers.
To simplify routine clinical practice and mitigate absolute quantification reliability issues between different systems, PSMA expression in a site of local or metastatic disease can be defined according to intensity in relation to the uptake in normal organs. This alternative strategy for PET quantification has been successfully applied in neuroendocrine tumors, with the Krenning score defining somatostatin receptor expression [40], and in Lymphoma response assessment using the 5-point scale Deauville criteria [41]. Our group has also described a scoring system for 18 F-Fluorthymidine [42]. Recently, a similarly pragmatic approach was used in a phase-II study evaluating the efficacy of 177 Lu-PSMA-617 in men with metastatic castration-resistant PC [29], in which patients were deemed suitable for therapy when lesional 68 Ga-PSMA-11 uptake was significantly greater than normal liver. Therefore, it is particularly important to access liver uptake agreement between different PSMA tracers. This study demonstrated an acceptable quantitative liver uptake agreement between 18 F-DCFPyL and 68 Ga-PSMA-11. However, a weak positive correlation between liver uptake values and 18 F-DCFPyL uptake time was found and, accordingly, in patients where 18 F-DCFPyL scans were performed at later time-points there was a significantly larger bias, in which liver uptake was higher than the corresponding scans using 68 Ga-PSMA-11. Although the optimal time point for 18 F-DCFPyL imaging is still not fully Moreover, the intrinsic variability of the tracers in normal organs must be well understood in order to be able to attribute tumor signal alterations to real changes in tumor mass, disease progression or response to treatment. The liver was the solid organ with the lowest COVs for both tracers, validating it as an appropriate reference tissue for thresholding images when assessing serial scans. Appropriate image thresholding is vital for qualitative assessment of PET scans [43]. We found a wider range of uptake values within the liver for 18 F-DCFPyL scans than those observed in another series of the literature where the SUV mean COV was 13.8% [36], but still acceptable in relation to 18 F-FDG (SUL mean COV 21.0-23.1%) [44]. This probably reflects the wider range of 18 F-DCFPyL uptake time in this series and a lower population number and may have contributed in some extent to the observed differences between 68 Ga-PSMA and 18 F-DCFPyL.
There were some limitations to this study. The cohort was relatively small, included patients undergoing imaging for different indications and there was a non-negligible period between 68 Ga-PSMA-11 and 18 F-DCFPyL scans. We minimized possible systematic errors by applying stringent (albeit arbitrary) inclusion criteria. As PSMA-avid tumor burden significantly correlates to PSA levels [24,45], we excluded patients with high PSA at either scan or significant interval biochemical progression, thus ensuring normal-organ comparability. Furthermore, even with rigorous calibration, the use of 4 different PET/ CT systems may influence data output. The retrospective design also impaired optimal protocol coincidence using the two tracers. The mean injected dose and uptake time after tracer injection were higher for 18 F-DCFPyL, but this variation was considered in the applied statistical analysis and also reflects routine clinical practice since, generally, lower doses and narrower acquisition periods are used for 68 Ga-labeled tracers. While the optimal time point for 18 F-DCFPyL is not yet fully established, preliminary data support imaging at 120 min post injection [38]. Accordingly, our 18 F-DCFPyL protocol has since evolved to imaging at a later time-point, resulting in a somewhat heterogeneous acquisition time-frame for this population. We acknowledged this issue and performed subgroup analysis of physiologic uptake at different time-points. Finally, although there is published data on the change in PSMA expression in tumor lesions following androgen deprivation therapy [46,47], the effect on normal organ distribution is still unclear. Noticeably, only 3 (8.8%) patients had a systemic treatment commenced and one (2.9%) patient had it stopped in between scans.
This study is not intended to provide a comparison of lesion sensitivity, but there is preliminary, intra-patient comparison data [18,20]. Head-to-head comparisons between these agents are required to define their relative diagnostic performance but these data further encourage evaluation of 18 F-DCFPyL given the advantages it provides in terms of radiopharmaceutical supply.

Conclusion
Normal tissue biodistribution patterns of 68 Ga-PSMA-11 and 18 F-DCFPyL were similar, despite subtle differences in quantitative values. Liver uptake demonstrated an acceptable agreement and low inter-patient variability between the two tracers, allowing its use as a reference organ for thresholding scans for qualitative comparison of PSMA expression when using these different tracers.

Additional file
Additional file 1: Table S1. Correlation between clinical/protocol variables and each tracers' quantitative uptake in the target organs. P values in bold reflect statistical significance. Table S2