Introduction
Prostate cancer (PCa) is the second most common cancer amongst men in the world, as recorded in 2018 [
1]. Molecular imaging of this malignancy either in primary or metastatic setting is presently dominated by the ligands directed to the prostate-specific membrane antigen (PSMA). This is a membrane-bound enzyme which is overexpressed in PCa cells compared to benign prostatic tissue by approximately 100- to 1000-fold [
2,
3].
68Gallium-labeled PSMA compounds, such as [
68Ga]Ga-PSMA-11, is therefore considered a highly tumor-specific radiotracer for PCa. Since PSMA agents have an unprecedented accuracy in recurrent PCa, it has been rapidly adopted in the clinic over the last years [
4,
5]. In staging of primary PCa, Hofman et al
. recently published a prospective study, showing the higher diagnostic accuracy of [
68Ga]Ga-PSMA-11 PET/CT in men with high-risk primary prostate cancer, as compared to conventional imaging (CT and bone scan) [
6], which is as well supported by retrospective studies [
7‐
9].
In many solid tumors, active surveillance and response monitoring with
18fluorine-fludeoxyglucose (2- [
18F]FDG) PET/CT is quite common, and adopted in various guidelines [
10]. Variations in [
18F]FDG-accumulation can provide valuable information on the activity and efficacy of new cancer therapeutics. The Positron Emission Tomography Response Criteria in Solid Tumors (PERCIST) or European Organization for Research and Treatment of Cancer (EORTC) criteria are often used to quantify the response on therapy using [
18F]FDG PET/CT. These criteria classify the disease status as ‘responder’, ‘progressed’, or ‘stable’ based on changes in the semiquantitative standard uptake values (SUV), corrected for lean body mass (SUL) [
10]. However, [
18F]FDG PET/CT is usually not suitable in PCa, as most tumors show limited FDG-accumulation, especially in hormone-sensitive setting.
Monitoring the response after therapy with [
68Ga]Ga-PSMA-11 PET/CT may be helpful in PCa, yet this approach is not validated yet [
11]. As with FDG-accumulation, the extent and intensity of the PSMA-uptake can be compared between scans to quantify response, and when deemed necessary adjust therapy. However, not many papers are published about response monitoring or active surveillance using [
68Ga]Ga-PSMA-11 PET/CT [
12]. One study by Gupta et al. [
13] compared the functional criteria PERCIST 1.0 and EORTC in [
68Ga]Ga-PSMA-11 PET/CT with the morphological criteria according to RECIST V1.1 in patients with metastatic PCa and biochemical progression. According to this study, molecular imaging criteria performed best in detecting progression based on changes of ≥ 25% SUV
mean (EORTC) or ≥ 30% SUL
peak (PERCIST) after hormone treatment [
13]. However, the study does not describe if the biological variation in [
68Ga]Ga-PSMA-11 uptake is comparable to that of [
18F]FDG, and if the same cutoff values apply, while only changes that exceed the normal variability should be interpreted as treatment response or disease progression
. Although two additional studies assessed this test–retest repeatability in metastatic prostate cancer using [
68Ga]Ga-PSMA-11 [
14] and [
18F]DCFPyL [
15], no studies in the primary setting are performed to this day, as far as we are aware of. Therefore, the aim of this study was to assess the day-to-day variability of [
68Ga]Ga-PSMA-11 uptake and visual interpretation in patients with primary prostate cancer.
Discussion
To our knowledge, repeatability of [68Ga]Ga-PSMA-11 PET/CT scans in the primary PCa setting has not been investigated before. This information is essential to perform response monitoring based on PSMA PET/CT. Therefore, the aim of this study was to assess the repeatability of [68Ga]Ga-PSMA-11 PET/CT scan in primary PCa patients in a 4-week interval.
The repeatability coefficient of SULpeak in the prostate tumor was 18.1%, suggesting that below this value, the absolute difference between two scans in one patient (under the same circumstances) should fall within 95% probability. If the relative difference in SULpeak is larger than ± 18%, then the difference is more likely to be explained by true changes and then by measurement errors. The absolute RC in the prostate tumor is probably less useful than the percentage, as there is a broad range in uptake between patients. The SULmean of normal tissue has a RC of 23.8% on average. In general, the RC of the SULmax is higher than the RC of SULmean, since SULmax is more susceptible to noa. The RC of the SULmean of the blood pool was 26%. However, due to the very low radiotracer activity in the bone and blood pool, small absolute differences in SULmean can already lead to a large percentage difference. Thus, in these organs, it is more informative to look at the repeatability coefficient of the absolute difference, which is 0.2 for the blood pool and 0.2 for bone.
The visual differences between the two scans were not considered clinically relevant by the expert readers, and were predominantly related to a variable urinary excretion or variances in noise levels. Particularly, in three patients the activity in the prostate was noisy; therefore, the primary tumor was difficult to distinguish from the background, probably explaining the initial inter-observer differences in visual assessment. Next, in one of the patients, the visual appearance of a lymph node was different, which might have been caused by a larger (22%) deviation in the injected dose between the two scan points in that particular case. Still, in this patient the tumor and other organs were not visually different.
The current study had quite a similar setup as previous studies described in the literature, but there are some relevant differences [
14,
15]. Previous studies looked at metastatic patients or [
18F]DCFPyL instead of [
68Ga]Ga-PSMA-11. Given the similar bio-distribution, findings for [
68Ga]Ga-PSMA-11 may probably be generalized to [
18F]DCFPyL [
22]. Noteworthy is that [
18F]PSMA-1007 has a distinctly different bio-distribution than the latter radiotracers [
23]. Also, most other studies quantified uptake with body-weight-corrected SUV instead of SUL, yet the RC of SUL
max and SUV
max within the same patient in a test–retest setting is comparable. The RC of SUL
max of the prostate lesion in our study was 21.9%. Pollard et al
. reported 30.3% and Jansen et al
. 31.0% SUV
max. The wCV of the prostate tumor in our study is 7.9% for SUL
max, compared to 10.9% for SUV
max reported in the study by Pollard et al. [
24]. These studies investigated patients with metastatic lesions as opposed to the patients with primary prostate cancer in our study. Metastatic lesions have a larger variation in PSMA receptor expression and general expansion of disease between patients than primary PCa tumors, thus possibly resulting in a higher RC than ours. When comparing our TBR findings to Jansen et al
. [
15], the RC is within the same range (38.4% vs 37.3%). The RC of the SUL is smaller than the RC of the TBR (21.9% vs 38.4% SUL
max vs TBR
max), probably because TBR adds the SUL variation both the blood pool and the tumor. This was found by Jansen et al
. as well [
15]. In prior studies, a shorter window of not more than 7 days for repeatability was used. Due to the fact that prostate cancer is generally an indolent cancer, a 4-week interval was considered reliable, as shown in the present study.
In general, the SUL
max values of normal tissues found in the literature for PSMA PET/CT scans are comparable to ours [
16,
25‐
27]. The differences between the two scans are largest in the ureter and bladder, which can be explained by variations in urine volume and radiotracer activity in the bladder. Though patients were not allowed to receive furosemide according to study protocol study, two patients had a protocol violation and received furosemide before one acquisition, resulting in a clear difference in bladder radiotracer activity (i.e., 2.1 vs 9.6 SUL
mean). The differences between two scans are lowest in the liver, concordant to the findings of Li et al
. [
28]. The uptake variation of the parotid is on average repeatable (23.4%), although there is a large range, and comparable to the results of Pollard et al. (26.5%). However, comparing the SUL
mean to [
18F]DCFPyL found by Li et al. [
28], the parotid and kidney SUL
mean were lower than our findings. Still SUL
mean of the liver was equal.
The repeatability of [
68Ga]Ga-PSMA-11 uptake in prostatic lesions is not entirely comparable to previous findings of [
18F]FDG uptake in malignancies [
21]. However, [
18F]FDG scans do require a more concise preparation and patients’ adherence to the protocol [
29]. In the day-to-day clinical setting, patients are likely to have variation in, for instance, blood glucose levels or physical activity, thus directly resulting in a less reproducible [
18F]FDG uptake. The advantage of [
68Ga]Ga-PSMA-11 PET/CT is that signal intensity depends primarily on the number of PSMA-expressing cells and expression density per cell [
30]. In contrast to [
18F]FDG, which accumulates to some extent in most tissues in the body, PSMA-ligands tend to accumulate only in specific tissues. Sahakyan et al. [
31] found that variability in the liver and kidneys can be caused by intrapatient factors (i.e., time of day, recent meals, hydration status), and that interpatient factors (i.e., weight, height, body composition, medical comorbidities) can influence uptake in the salivary glands. Nonetheless, these influences were described in patients who underwent therapy, so it is difficult to distinguish between therapeutic effects and day-to-day physiological variations.
All these characteristics aid in repeatable quantitative [
68Ga]Ga-PSMA-11 uptake in PCa lesions, and so it should allow for stable follow-up monitoring of the disease. A change in PSMA uptake in the tumor more than 18% as mentioned before may indicate either disease progression or treatment response. In a preclinical setting, PSMA expression is already used for response monitoring in taxane-based chemotherapy [
32]. Note that careful image interpretation is needed when describing an increase or decrease in PSMA uptake [
33]. Androgen deprivation therapy can influence the PSMA expression, where up- or downregulation is not unambiguously affected by type and duration of medication [
34‐
36]. If PSMA PET/CT is used for response monitoring of radionuclide therapy, the tumor sink effect cannot be neglected [
37].
Limitations
Our study has some limitations that need some further elaborations. First, there was an alteration in the clinical imaging protocol while performing this study. The prescribed activity of [
68Ga]Ga-PSMA-11 was increased from 100 MBq (
n = 20) to 150 MBq (
n = 2) in order to improve the image quality of PET/CT images in our institute. Though the effects on the SUL
mean and SUL
peak RC-values are limited (Additional file
1: Figure 1), the effect is somewhat larger in SUL
max measurements, as these are more susceptible to noise. Other aspects of the protocol like tracer uptake time, use of furosemide, and reconstruction settings were not altered. Although stringent protocol adherence was aimed for, four patients violated the ± 10% variation in administered dose and ± 5 min variation in uptake time. As this was not an exclusion criterion and might also occur in clinical practice as well, we decided to include these patients. This, however, did result in an increased repeatability compared to the use of a clean dataset without these violations (Additional file
1: Figure 2). Still, if centers want to perform response assessment with PSMA-PET/CT, it remains important to adhere to the protocol.
For image processing, VOIs were used instead of segmenting the entire organs to provide SUL
mean values. Still, Li et al. [
27] found that there was no significant difference in SUL
mean of an entire segmented liver and a VOI with a 3 cm in diameter. In addition, images were not registered to each other before segmentation, and the VOIs were drawn by one person based on visual concordance of the location in the two PET/CTs. We believe that registration is not commonly used in clinical practice, as it might induce registration errors, and that the chosen segmentation approach mimics our clinical routine for response monitoring.
Next, we did not report on SULmean repeatability of the prostate tumor, as threshold-based segmentations using, for instance, 45% of SULmax proved not an appropriate method, especially for tumors with low SULmax. In these patients, almost the entire prostate was segmented with this threshold, thus not representing the actual prostate tumor. To our knowledge, no standardized methods are yet published to accurately define the prostate lesion volume based on PSMA PET/CT. Since SULmax is more sensitive to image noise, we chose to obtain the SULpeak for prostate lesions.
Publisher's Note
Springer Nature remains neutral with regard to jurisdictional claims in published maps and institutional affiliations.