**2. Materials and Methods**

#### *2.1. Study Design*

Our investigation included 88 consecutive patients with elevated serum PSA levels and with biopsy-confirmed PC who underwent PSMA PET/CT for primary staging and specifically for the detection of possible metastases. For the retrospective analysis of the data, the datasets of patients who had received prior prostate therapy were excluded. The data for the period 2017 to 2021 were collected at a practice for Radiology and Nuclear Medicine in Cologne, Germany. Fifty-two patients underwent 18F-PSMA, and thirty-six patients underwent 68Ga-PSMA. The PSMA uptake of the 18F-PSMA and of the 68Ga-PSMA PET findings were quantified as SUVmax. The PSMA-positive lesions in the included patients were compared with histopathologic results of the prostate biopsies. A prostate biopsy was performed in all patients. PC was verified histologically with TRUS-guided or multiparametric MRI (mpMRI)-fusion guided prostate biopsy. In all patients, an adenocarcinoma of the prostate was histopathologically proven by biopsy. The biopsy results expressed as Gleason Score (GS) formed the reference basis for the PSMA PET/CT findings. Clinically significant PC (csPC) was defined as GS 7b-tumors or greater (any ISUP grade group ≥ 3) (subgroup: csPCa) and as GS 8-tumors or greater (any ISUP grade group ≥ 4) (subgroup: csPCb) [2].

#### *2.2. Positron Emission Tomography/Computed Tomography Imaging Protocol and Interpretation*

The study was performed using a PET/CT scanner (Gemini TF16; Philips Medical Systems, Best, The Netherlands). PET/CT images were acquired in 3D acquisition mode (matrix 168 × 168) 90 ± 10 min after intravenous injection of 326 ± 51.8 MBq [18F]PSMA-1007 or 60 ± 10 min post injectionem (p.i.) of 257 ± 85.7 MBq [68Ga]Ga-PSMA-11. PET images from the skull base to the proximal thigh were acquired for 3 min per bed position (axial field of view: 21.8 cm). A maximum inspiratory contrast-enhanced CT in the venous phase was performed in all included patients for attenuation correction and anatomical correlation. Decay, random, scatter, and attenuation correction were implemented. PET image reconstruction was carried out by using an ordered-subset expectation maximization (OSEM)-algorithm with 2 iterations and 14 subsets and Gaussian filtering with 4.2 mm transaxial resolution at full width at half maximum. Volumes of interest (VOIs) were drawn

on the foci suspected of being malignant due to the PSMA distribution pattern on PET in consensus with CT imaging. Values for tracer uptake expressed as the SUVmax measured on the VOIs were plotted on a receiver operating characteristic (ROC) curve. The area under the ROC (AUC) as well as the best cut-off level for SUVmax to classify the VOIs were calculated. Two experienced board-certified nuclear medicine physicians and two experienced board-certified radiology physicians, each of them with more than 5 years of experience in PSMA PET/CT hybrid imaging, assessed the images by consensus.

### *2.3. Statistical Analysis*

Numeric data are presented as median or mean ± standard deviation (SD). We evaluated the relationship between PSMA PET/CT positivity (e.g., expressed as SUVmax) and clinical parameters such as GS. To compare the two patient cohorts 18F-PSMA and 68Ga-PSMA and identify differences between them, we performed Student's *t*-tests for data that showed a normal distribution or nonparametric Mann–Whitney U tests for sample data that was not normally distributed. Using a ROC curve analyses, the performances of the procedures (18F-PSMA and 68Ga-PSMA) for distinguishing between PC with lowand intermediate-favorable risk vs. intermediate-unfavorable and high-risk as well as between low- and intermediate-risk vs. high-risk were calculated by plotting sensitivity against 1-specificity. Optimal SUVmax cut-off values were rated using the Youden index for the separate methods (18F-PSMA and 68Ga-PSMA). A *p* value < 0.05 was considered as statistically significant. We carried out the statistical analyses using SPSS version 27.0 (IBM SPSS Statistics Corporation, Ehningen, Germany).

### **3. Results**

We identified 88 patients who underwent 18F-PSMA (52) or 68Ga-PSMA (36). The median age was 67.5 years (range 51–80 years) in the patient group of 18F-PSMA and 65.5 years (range 48–79 years) in patients whose imaging was conducted with 68Ga-PSMA. Clinical and pathological characteristics of the study population are summarized in Table 1.


#### **Table 1.** Patients' characteristics.

Abbreviations: PSMA, prostate-specific membrane antigen; PET/CT, positron emission tomography/computed tomography; SD, standard deviation; y, year; PSA, prostate-specific antigen.

PSMA-avid lesions were found in all 52 study patients in the 18F-PSMA cohort and in 97.2% (35/36) of the 68Ga-PSMA cohort. The 18F-PSMA scans detected prostatic lesions with elevated PSMA avidity in 100% (52/52), LNM in 32.7% (17/52), and bone metastases in 17.3% (9/52) of cases. A total of 35 out of 36 (97.2%) untreated patients, who underwent a 68Ga-PSMA, showed lesions with an elevated tracer uptake in the prostate. 68Ga-PSMA scans also detected LNM in 16.7% (6/36) and bone metastases in 8.4% (3/36) of cases. A total of 33 patients, who underwent 18F-PSMA, demonstrated solitary PSMA tracer-positive prostatic lesions, whereas 26 patients showed them in the 68Ga-PSMA group.

In our study, 5.8% (3/52) of PSMA-positive PET lesions, based on all patients with PSMA-positive findings, in the 18F-PSMA cohort and 11.4% (4/35) in the 68Ga-PSMA cohort were categorized as low-risk PC (GS < 7) with ISUP grade group 1. In one patient with a biopsy finding of GS 6, no increased PSMA avidity was detected in the PET/CT with [ 68Ga]Ga-PSMA-11. Intermediate-risk PC (GS 7) with ISUP grade groups 2 and 3 occurred in 40.4% (21/52) of 18F-PSMA-positive and in 60% (21/35) of 68Ga-PSMA-positive patients, whereas 53.8% (28/52, 18F-PSMA) and 28.6% (10/35, 68Ga-PSMA) showed high-risk PC lesions with an ISUP grade group 4 to 5 (Table 2, Figures 1 and 2).


**Table 2.** PSMA-positive scan lesions for staging patients in relation to the Gleason Score (GS).

\* Fisher exact test. Abbreviations: PSMA, prostate-specific membrane antigen; PET/CT, positron emission tomography/computed tomography; LNM, Lymph node metastases; GS, Gleason Score; *p* < 0.05 is considered significant; *r*, Pearson correlation coefficient.

The 88 study patients were separately (18F-PSMA and 68Ga-PSMA) grouped into categories by GS and compared as follows: patients with GS 6 and GS 7 vs. patients with GS ≥ 8 and with GS 6 and GS 7a vs. patients with GS ≥ 7b (Figures 1 and 2).

In the 18F-PSMA cohort, PC prostatic lesions with histopathology of low- and intermediatefavorable risk PC (GS ≤ 7a) were shown in 25% (13/52) compared to 75% (39/52) with histopathology of intermediate-unfavorable and high-risk PC (GS ≥ 7b) (Figure 1). PSMAavid metastases and PSMA-positive LNM were shown in 5.8% with GS ≤ 7a (3/52) vs. 30.8% with GS ≥ 7b (16/52) and in 3.8% with GS ≤ 7a (2/52) vs. 28.8% (15/52) with GS ≥ 7b (Figure 1).

For the 68Ga-PSMA cohort, the distribution of the PSMA-avid PC lesions in the prostate was as follows: 31.4% (11/35) with GS ≤ 7a vs. 68.6% (24/35) with GS ≥ 7b, respectively (Figure 2). Neither PSMA-positive metastases nor LNM were shown in the subgroup with GS ≤ 7a, whereas the subgroup with GS ≥ 7b revealed PSMA-avid metastases in 25.7% (9/35) of cases and positive LNM in 17.1% (6/35) (Figure 2).

**Figure 1.** Distribution of positive findings (shown by [18F]PSMA-1007 PET/CT) classified by Gleason Score (GS) categories (patients with GS < 7 to GS ≥ 8 and the comparison of GS ≤ 7a versus patients with GS ≥ 7b).

**Figure 2.** Distribution of positive findings (shown by [68Ga]Ga-PSMA-11 PET/CT) classified by Gleason Score (GS) categories (patients with GS < 7 to GS ≥ 8 and the comparison of GS ≤ 7a versus patients with GS ≥ 7b).

The PSMA uptake of the [18F]PSMA-1007 and of the [68Ga]Ga-PSMA-11 PET findings was quantified as SUVmax. Comparing 18F-PSMA and 68Ga-PSMA scanned patients, there was no statistical significance for the differentiation of mean and median SUVmax for the most intense prostatic lesions (*p* = 0.224) (mean SUVmax ± SD: 12.2 ± 10.4 vs. 10.0 ± 8.0, median SUVmax 9.0 vs. 6.7).

When using a SUVmax of 2.5 as the cut-off value between PC lesions in the prostate with low- and intermediate-favorable risk (GS ≤ 7a) vs. with intermediate-unfavorable and high-risk (GS ≥ 7b), 18F-PSMA indicated a sensitivity of 100%, a positive predictive value (PPV) of 76%, and an accuracy of 76% (Table 3). For 68Ga-PSMA, the sensitivity was 97%, the PPV was 75%, and the accuracy was 77%, respectively (Table 3).


**Table 3.** Test parameters for the staging of prostate cancer with [18F]PSMA-1007 PET/CT and with [ 68Ga]Ga-PSMA-11 PET/CT; distribution of positive prostatic findings classified by Gleason Score (GS) categories (GS ≤ 7a versus GS ≥ 7b; GS ≤ 7 versus GS ≥ 8).

Abbreviations: SUVmax, maximum standardized uptake value; vs., versus; PSMA, prostate-specific membrane antigen; PET/CT, positron emission tomography/computed tomography; NPV, negative predictive value; PPV, positive predictive value. \* for [18F]PSMA-1007 PET/CT SUVmax 8.95, [68Ga]Ga-PSMA-11 PET/CT SUVmax 8.7. \*\* for [18F]PSMA-1007 PET/CT SUVmax 4.75, [68Ga]Ga-PSMA-11 PET/CT SUVmax 6.2.

Using the Youden index, the best analyzed cut-off value for 18F-PSMA was a SUVmax of 8.95 (subgroup: 18F-7a/b) for distinguishing GS ≤ 7a from GS ≥ 7b prostatic lesions. ROC analysis showed an AUC of 0.750 (95% Cl 0.590; 0.911; SD (AUC) = 0.082; *p* = 0.007) for the comparison with a SUVmax of 8.95 (18F-7a/b). The sensitivity, the specificity, the PPV, and the accuracy for 18F-7a/b was 62%, 85%, 92%, and 67%, respectively. For the differentiation of GS ≤ 7 from GS ≥ 8 (subgroup: 18F-7/8) an AUC of 0.592 (95% Cl 0.539; 0.881; SD (AUC) = 0.055; *p* = 0.26) with a SUVmax of 4.75 (18F-7/8) was evaluated with a sensitivity of 90%, a specificity of 52%, a PPV of 61%, and an accuracy of 63%, respectively (Table 3).

By means of ROC analysis, the best cut-off value for 68Ga-PSMA was a SUVmax of 8.7 (subgroup: 68Ga-7a/b) to differentiate GS ≤ 7a and GS ≥ 7b PC lesions (AUC = 0.814; 95% Cl 0.668; 0.961; SD (AUC) = 0.075; *p* = 0.003) with a sensitivity of 54%, a specificity of 91%, a PPV of 93%, and an accuracy of 66%. The best AUC for distinguishing GS ≤ 7 from GS ≥ 8 PC lesions was 0.710 (95% Cl 0.539; 0.881; SD (AUC) = 0.087; *p* = 0.055) with a SUVmax of 6.2 (subgroup: 68Ga-7/8) and with a sensitivity of 89%, a specificity of 33%, a PPV of 43%, and an accuracy of 63% (Table 3).

Figure 3 shows a 18F-PSMA with a histopathologically confirmed aggressive PC with a GS of 8 (4 + 4), without locoregional LNM and without skeletal metastases, but with three mediastinal LNM of normal size, located infracarinally and bilaterally hilar with a high PSMA avidity, and Figure 4 shows a 68Ga-PSMA with a histopathologically confirmed aggressive PC with a GS of 8 (4 + 4) with locoregional LNM and without distant LNM and without skeletal metastases.

**Figure 3.** *Cont*.

**Figure 3.** (**a**–**c**) Case study of a patient with evidence of a prostate-specific membrane antigen (PSMA) avid prostatic finding in the initial staging (**a**) (with an initial prostate-specific antigen of 5.0 ng/mL) concordant with the histopathologically confirmed aggressive prostate carcinoma (PC) with a Gleason Score of 8 (4 + 4). The [18F]PSMA-1007 positron emission tomography/computed tomography showed no locoregional lymph node metastases or skeletal metastases, but three mediastinal lymph nodes ((**a**): blue arrows) of normal size, located infracarinally (**b**) and bilaterally hilar ((**c**): hilar right), carrying intensive tracer uptake (the highest maximum standardized uptake value of 11.4), which were histopathologically confirmed as metastatic PC.

(b)

(c)

**Figure 4.** Case study of a patient with evidence of a prostate-specific membrane antigen (PSMA) avid prostatic finding in the initial staging with an initial prostate-specific antigen of 13.0 ng/mL, concordant with the histopathologically confirmed aggressive prostate carcinoma (PC) with a Gleason Score of 8 (4 + 4). The [68Ga]Ga-PSMA-11 positron emission tomography/computed tomography showed five locoregional lymph node metastases (blue arrows) carrying intensive tracer uptake (the highest maximum standardized uptake value of 19.4), which were histopathologically confirmed as metastatic PC.

#### **4. Discussion**

The EAU-EANM-ESTRO-ESUR-ISUP–SIOG Guidelines 2022 explicitly emphasize that most published studies on the primary staging of PC were based on 68Ga-labeling for PSMA PET imaging, and few studies were based on 18F labeling [2,7]. According to these guidelines, there are currently no conclusive data comparing 68Ga-PSMA with 18F-PSMA imaging in primary PC staging. In this context, the present study can possibly make a valuable contribution to the comparison of the two methods, 68Ga-PSMA and 18F-PSMA, in the clinical staging of PC.

In this comparative study of 68Ga-PSMA vs. 18F-PSMA in patients with newly diagnosed PC, we analyzed the PSMA-positive lesions that were determined to be malignant. PSMA-avid prostatic foci in concordance with histopathologically proven PC were found in all 52 study patients in the 18F-PSMA cohort, while 68Ga-PSMA showed them in 97.2% of the cohort (35/36). The imaging data for prostatic lesions were compared with histopathologic prostate biopsy results expressed as GS. Our results showed concordant findings with both tracers, which is in line with other studies comparing 18F-PSMA and 68Ga-PSMA in primary staging [8–10]. Kuten et al. reported in a head-to-head comparison that the identification of all intermediate- and high-risk PC lesions was comparable by both methods [8]. Hoberück et al. described, in a retrospective intraindividual comparison, that 18F- as well as 68Ga-PSMA appeared largely interchangeable, with neither tracer significantly outperforming the other [9]. The authors described that no significant difference considering SUVmax of tumor lesions was shown [9]. A prospective intraindividual comparative study on 18F-PSMA and 68Ga-PSMA for PC staging, evaluation at biochemical recurrence and assessment of metastatic disease, by Pattison et al. demonstrated a high concordance of 92% for TNM stage [10]. Further studies confirmed similar findings in PSMA PET/CT imaging with the two radiopharmaceuticals in the setting of restaging PC patients, too [11,12]. Rauscher et al. showed similar detection rates in patients with biochemical recurrence after radical prostatectomy. However, five times as many positive findings of benign origin were found in 18F-PSMA compared with 68Ga-PSMA [11]. The side-by-side evaluation specifically requested by the authors for the 18F-PSMA diagnosis of PET and CT images as well as intensive reader training on well-known pitfalls (for example, non-specific tracer uptake in the ganglia) in the clinical context [11] was implemented in a quality-assured manner by the diagnostic specialists in our present study. In a further restaging study by Hoffmann et al., both methods (18F-PSMA and 68Ga-PSMA) showed comparable overall findings [12]. Exceptions to this, however, were a clearer distinction between positive and negative results in the 18F-PSMA imaging considering a PSA threshold, determined in the study, in biochemical recurrent patients after radical prostatectomy [12]. However, Rahbar et al. described on the basis of patient images that 18F-PSMA offers an advantage over imaging with 68Ga-PSMA for the detection of local recurrence after primary local therapy due to the later renal tracer excretion. The authors related this advantage to case constellations with unclear lesions near the ureter or the urinary bladder [13]. Renal excretion of 68Ga-PSMA and radioactive bladder filling obscures local recurrence in the situation of biochemical recurrence but is of less relevance in initial tumor staging as in our study. Considering the comparison of 68Ga-PSMA and the PET/CT with another 18F-labeled radiotracer, named [18F]rhPSMA-7 (18F-rhPSMA-7), a study by Kroenke et al. showed similar tumor positivity rates and SUVmax values for primary PC and biochemical recurrence of PC [7,14]. Giesel conducted a comparative study considering different 18F-labeled PSMA PET ligands. The comparison of [18F]DCFPyl PET/CT (18F-DCFPyl) with 18F-PSMA also showed no significant differences in the detection of carcinoma foci or their SUVmax values [6].

In order to improve underdetection of high-grade PC and overdetection of low-grade PC [2,4], it makes sense to define a separation sharpness for the clinical setting. The cancer patients who would not benefit from a therapy should be considered separately from the patients with expected therapy success. The EAU-EANM-ESTRO-ESUR-ISUP–SIOG Guidelines 2022 do not specify how the term csPC should be defined exactly [2]. The guidelines report that studies mostly define GS 7 tumors and upwards or GS 7b tumors and upwards as clinically significant and that authors should decide for themselves and explain this in the study design [2]. In our study, we defined in one patient subgroup csPCa as any ISUP grade group ≥ 3 malignancy (patients with the high–intermediate or intermediateunfavorable PC risk of GS 7b and above) and in a second patient subgroup csPCb as any ISUP grade group ≥ 4 malignancy (patients with the high PC risk of GS 8 and above), in order to then be able to compare both groups. Our study mainly focused on analyzing the best SUVmax cut-off value to identify the clinically significant PC foci and to compare the results of both methods. PSMA-avid lesions were defined as suspicious of malignancy when the uptake of the tracer was significantly higher than the surrounding benign tissue, when the tracer uptake appeared focal in character, and when the lesions were classified as primarily malignant (in the opinion of experts based on their extensive experience in the interpretation of PSMA PET/CT scans). Experience has shown that suspicious PET lesions with a SUVmax of 2.5 or higher were mostly associated with compatible and duplicatable visual evidence of PC foci and, therefore, this value was initially used as a cut-off to distinguish between PET positivity and negativity for both radiopharmaceuticals. Because the tumor-to-background ratio for the malignant lesions compared with the benign tissue in the PSMA PET/CT is very high according to previous studies (e.g., in comparison to FDG PET/CT, [15]) and the difference in the detected lesions was clearly shown in the present study, we did not list the SUVmean values separately, as this would have no added value.

First, choosing a routinely used SUVmax of 2.5 as the cut-off value between csPC and clinically insignificant PC, the findings of both methods demonstrated similar concordance in our study. 18F-PSMA revealed 25% (13/52) of PC prostatic lesions with histopathology of low- and intermediate-favorable risk PC (GS ≤ 7a) vs. 75% (39/52) with histopathology of intermediate-unfavorable and high-risk PC (GS ≥ 7b) with a sensitivity of 100%, a PPV of 76%, and an accuracy of 76% considering a SUVmax of 2.5. For 68Ga-PSMA, the results were 31.4% (11/35) vs. 68.6% (24/35) with a sensitivity of 97%, a PPV of 75%, and an accuracy of 77% with the uptake of the radiotracer above a SUVmax of 2.5. In the present study, because the specificity of both methods was extremely low (10% vs. 27%) using a SUVmax threshold of 2.5, an optimal SUVmax cut-off value was determined for 18F-PSMA and for 68Ga-PSMA by Youden index calculation. The reasons for reduced specificity in PSMA imaging are well known and include neovascularization and PSMA overexpression in non-prostatic tissue, e.g., benign neoplasms, i.e., thyroid and parathyroid adenomas, and in non-prostatic malignancies such as breast cancer, thyroid cancer, gliomas, lung cancer, neuroendocrine tumors, lymphoma, and renal cell carcinoma. There are fewer false positives if the PSMA images are interpreted by experts who are aware of the various pitfalls [16].

Subsequently, ROC curves were used to characterize the diagnostic performance. By considering the PSMA-avid prostatic lesions and the corresponding classification in the GS based on the biopsy, a SUVmax of 8.95 was analyzed by ROC analysis (*p* = 0.007) to differentiate between csPC and clinically insignificant PC (subgroup: csPCa) for 18F-PMSA with a sensitivity of 62%, a specificity of 85%, a PPV of 92%, and an accuracy of 67%. 68Ga-PSMA gave similar findings for a SUVmax of 8.7 (*p* = 0.003) with a sensitivity of 54%, a specificity of 91%, a PPV of 93%, and an accuracy of 66%, respectively. However, our data show a higher (but also moderate) specificity and a higher PPV for 18F-PSMA (52% and 61% based on a SUVmax of 4.75) in comparison with 68Ga-PSMA (33% and 43% based on a SUVmax of 6.2), when differentiating between low- and intermediate-risk PC vs. high-risk PC (subgroup: csPCb), with comparable sensitivity (90% vs. 89%) and accuracy (63% both). But these data did not show statistical significance (SUVmax of 4.75, *p* = 0.26 and SUVmax of 6.2, *p* = 0.055). Kuten et al. calculated ROC curves to distinguish pathological from non-pathological components of the prostate, for which both methods proved to be suitable [8]. A comparison of the results with our calculated values is not possible because the comparison groups differ. Additionally, due to the lack of statistical significance, no optimal SUVmax values could be calculated in the study by Kuten et al. [8].

The results of diagnostic PSMA imaging as part of the staging of PC offer the possibility of guiding biopsy and therapy management to detect the targeted PC lesions with the most aggressive tumor foci (csPC) [17,18]. A mpMRI in combination with a PSMA hybrid imaging fusion biopsy could increase the accuracy of directed biopsy [18]. Pepe et al. demonstrated a lower false positive rate and a better negative predictive value compared with mpMRI. In 80% of the cases, a biopsy could have been omitted based on the PSMA PET/CT results [18]. As part of individual therapy management, hybrid imaging with PSMA PET/CT enables optimal patient selection as well as personalized monitoring [17]. In this regard, our calculated SUVmax cut-offs can be used to differentiate between lowand intermediate-favorable from intermediate-unfavorable and high-risk PC lesions. The more we know about diagnostic imaging (such as the correlation between PSMA receptor density and GS as well as PSMA imaging with different radiopharmaceuticals and their physiological expression in non-prostatic benign tissue and non-prostatic tumors, both benign and malignant) and can optimize it, the better therapy decisions can be made [17]. Because present EAU Guidelines state that there is currently no conclusive data comparing 68Ga-PSMA vs. 18F-PSMA imaging in primary PC staging [2], we investigated this. The comparison could not show any clear advantage for one of the methods in our study, which is also an important statement for clinical application. In all 52 study patients in the 18F-PSMA cohort and in 97.2% (35/36) of the patients in the 68Ga-PSMA cohort, PSMA-avid prostate lesions were detected concordant with histopathologically proven PC. PSMA-positive metastases were shown in 5.8% (3/52) in the intermediate-favorable risk 18F-PSMA cohort vs. in 30.8% (16/52) in the intermediate-unfavorable risk group, but no PSMA-avid metastases (0/35) were seen in the 68Ga-PSMA intermediate-favorable risk cohort vs. 25.7% (9/35) with GS ≥ 7b. In view of the nearly similar results and the good performance of 18F- as well as 68Ga-labeled compounds, the challenge for the use of the appropriate radiopharmaceutical could potentially be made depending on availability [12]. Nevertheless, further studies are needed to assess the position of routinely established 68Ga- and 18F-labeled compounds in PSMA imaging and their actual clinical utility. These will be carried out on the different radiotracers in order to shed light on new aspects, the overall impact on survival, and the clinical impact of PSMA-based diagnostics such as PSMA-targeted biopsies [7,18]. In this context, a randomized study that would perform a combined PSMA imaging with a mpMRI as a guide for prostate biopsy in the initial stage with a high suspicion of csPC and would consider different radiotracers might be useful [19–22]. Limitations of the present study include the retrospective nature of the analysis, the small number of patients, and the lack of an intraindividual comparison of the patients. To confirm and expand our results we recommend further studies, ideally prospective with larger patient cohorts.
