Higher Preoperative Maximum Standardised Uptake Values (SUVmax) Are Associated with Higher Biochemical Recurrence Rates after Robot-Assisted Radical Prostatectomy for [68Ga]Ga-PSMA-11 and [18F]DCFPyL Positron Emission Tomography/Computed Tomography

This study aimed to investigate the association between the 68Ga- or 18F-radiolabeled prostate-specific membrane antigen (PSMA) tracer expression, represented by the maximum standardised uptake value (SUVmax) of the dominant intraprostatic lesion, and biochemical recurrence (BCR) in primary prostate cancer (PCa) patients prior to robot-assisted radical prostatectomy (RARP). This was a retrospective, multi-centre cohort study of 446 patients who underwent [68Ga]Ga-PSMA-11 (n = 238) or [18F]DCFPyL (n = 206) Positron Emission Tomography/Computed Tomography (PET/CT) imaging prior to RARP. SUVmax was measured in the dominant intraprostatic PCa lesions. [18F]DCFPyL patients were scanned 60 ([18F]DCFPyL-60; n = 106) or 120 ([18F]DCFPyL-120; n = 120) minutes post-injection of a radiotracer and were analysed separately. To normalise the data, SUVmax was log transformed for further analyses. During a median follow-up of 24 months, 141 (30.4%) patients experienced BCR. Log2SUVmax was a significant predictor for BCR (p < 0.001). In the multivariable analysis accounting for these preoperative variables: initial prostate-specific antigen (PSA), radiologic tumour stage (mT), the biopsy International Society of Urological Pathology grade group (bISUP) and the prostate imaging and reporting data system (PI-RADS), Log2SUVmax was found to be an independent predictor for BCR in [68Ga]Ga-PSMA-11 (HR 1.32, p = 0.04) and [18F]DCFPyL-120 PET/CT scans (HR 1.55, p = 0.04), but not in [18F]DCFPyL-60 ones (HR 0.92, p = 0.72). The PSMA expression of the dominant intraprostatic lesion proved to be an independent predictor for BCR in patients with primary PCa who underwent [68Ga]Ga-PSMA-11 or [18F]DCFPyL-120 PET/CT scans, but not in those who underwent [18F]DCFPyL-60 PET/CT scans.


Introduction
Prostate cancer (PCa) is the second most commonly diagnosed cancer among men around the world [1]. Due to the growing and increasingly aging population, a further rise in PCa cases can be expected [2]. This stresses the importance of accurate disease staging and the subsequently adequate treatment of PCa. Unfortunately, a substantial proportion of men experience the biochemical recurrence (BCR) of a disease after treatment with a curative intent [3]. In patients with localised PCa who underwent robot-assisted radical prostatectomy (RARP), BCR is seen in approximately 30% of patients within 10 years after surgery [4]. The current models predicting the risk of recurrence after surgery are based on both clinical and pathological variables, as well as on imaging techniques, such as multiparametric magnetic resonance imaging (mpMRI) [3,5].
Recently, Prostate-Specific Membrane Antigen Positron Emission Tomography/Computerized Tomography (PSMA PET/CT) was introduced as a novel imaging technique, improving the prediction of the disease outcome among patients with PCa compared to that of conventional imaging techniques [6]. In current practice, PSMA is usually labelled with 68 Gallium (e.g., [ 68 Ga]Ga-PSMA-11) or 18 Fluorine (e.g., [ 18 F]DCFPyL). PSMA PET/CT labelled with either 68 Gallium or 18 Fluorine has widely proven its prognostic value among patients with recurrent PCa and is therefore implemented in the follow-up of PCa [3,7]. However, there are some essential differences between the two tracers. The use of 18 Fluroine offers an improved PET image resolution compared to that of 68 Gallium thanks to its shorter positron range and higher positron yield [8]. Also, 18 Fluorine has a higher binding affinity to PSMA receptors [8], potentially enhancing the detection of small metastases. Moreover, 18 Fluorine is a cyclotron product with a longer half-life of 110 min compared to that of 68 Gallium, a generator product with a shorter half-life of 68 min. The cyclotrons on site enable the centralised and large-scale production of 18 Fluorine-radiolabelled PSMA tracers, potentially making them more cost-effective and suitable for clinical applications.
In line with the European Association of Nuclear Medicine (EANM) standardised reporting guidelines for PSMA-PET (E-PSMA), the semi-quantitative measurement of PSMA tracer uptake, expressed as maximum Standardised Uptake Values (SUV max ), is one important element when interpreting PET/CT scans [9,10]. Previous studies suggest that the SUV max of the primary prostate tumour may predict unfavourable disease outcomes, such as a higher pathological International Society of Urological Pathology Grade Group (pISUP), pathological tumour stage (pT) and lymph node involvement [11][12][13]. Furthermore, research showed that the SUV max in the primary staging of PCa represents an independent predictor for disease recurrence with the radiotracer [ 68 Ga]Ga-PSMA-11 [11,14,15]. However, the predictive value of PSMA expression with [ 18 F]DCFPyL for recurrent disease is yet unknown.
Therefore, the aim of this research was to investigate whether tumour PSMA expression, calculated as SUV max , on PSMA PET/CT for staging PCa contributes to the preoperative and postoperative predictions of BCR occurrence after curative treatment.

Study Design and Patient Population
This was a retrospective, multi-centre cohort study. Patients with histologically proven PCa who underwent RARP between August 2016 and June 2022 were included. All patients underwent [ 68 Ga]Ga-PSMA-11 PET/CT or [ 18 F]DCFPyL PSMA PET/CT prior to RARP, with or without extended pelvic lymph node dissection (ePLND). An ePLND was performed based on a ≥7% risk of lymph node involvement, as calculated using the Briganti 2012 nomogram or the Memorial Sloan Kettering Cancer Center (MSKCC) nomogram [16,17]. Patients were excluded if the preoperative PSMA PET/CT showed no focal PSMA uptake (non-PSMA-expressing tumour [miT0]) [10,18], if the serum PSA was not available postoperatively, if the patients underwent prior treatment for PCa, or if distant metastases were found on preoperative PSMA PET/CT. Patients were included in two reference centres of the Prostate Cancer Network Netherlands, the Antoni van Leeuwenhoek Hospital-Netherlands Cancer Institute (NCI) and the Amsterdam University Medical Centre (AUMC). This study was a secondary analysis of two studies that were approved by the local institutional review boards of the NCI and AUMC (institutional review numbers IRBdm19-348 and 2017.543, respectively). All patients signed informed consent forms when enrolled in the original studies, explicitly allowing secondary analysis of their study data.

Data and Outcome
The preoperative parameters that were collected included age, initial prostate-specific antigen (PSA) level, clinical tumour stage (cT), biopsy ISUP Grade Group (bISUP) and number of (positive) biopsy cores. Also, the prostate imaging and reporting data system (PI-RADS) score and radiological tumour stage (mT) were assessed using multiparametric magnetic resonance imaging (mpMRI). The intensity of PSMA expression, semiquantitatively expressed as the SUV max of the dominant intraprostatic lesion, and molecular imaging nodal stage (miN stage) were assessed using preoperative PSMA PET/CT, as well as the type of radiotracer used (i.e., [ 68 Ga]Ga-PSMA-11 or [ 18 F]DCFPyL). The postoperative variables that were collected included the pathological tumour stage (pT), pISUP, pathological nodal stage (pN stage) and surgical margin status. The outcome was biochemical recurrence, defined as a postoperative PSA level of 0.2 ng/mL or higher in the follow-up.  68 Ga]Ga-PSMA-11 was radiolabelled in-house using a fully automated system (Scintomics GmbH, Gräfelfing, Germany). At the NCI, PSMA PET/CT scans were performed 60 min post-injection (p.i.) of both 68 Ga-PSMA-11 and [ 18 F]DCFPyL PET/CT, while [ 18 F]DCFPyL PET/CT scans at the AUMC were performed 120 min p.i. [19]. All PET images were combined with either a low-dose CT (120-140 kV, 40-80 mAs) or a diagnostic CT scan (130 kV, 110 mAs) for anatomical correlation and attenuation correction. All PET images were taken according to the EARL standards and were corrected for scatter, decay and random coincidences [20].

Image Interpretation of PSMA-PET/CT Imaging
At both centres, [ 18 F]DCFPyL and [ 68 Ga]Ga-PSMA-11 PET/CT scans were assessed by experienced nuclear medicine physicians. The PSMA PET/CT scans were reported in accordance with the E-PSMA guidelines [10]. These reports included the primary intraprostatic lesion, and if present, secondary lesions, miT stage and the presence of lymph node (miN1/M1a stage), bone or visceral metastases (miM1b/c stage) [21]. Imaging results were primarily based on visual and semi-quantitative interpretations according to the E-PSMA guidelines [10]. Positive lesions were reported if the [ 68 Ga]Ga-PSMA-11/[ 18 F]DCFPyL uptake was higher than prostatic background activity (i.e., scores 4 and 5 based on a 5-point scale according to E-PSMA).

Scan Assessment and SUV max Assessment
For semi-quantitative analysis, the maximum standardised uptake value (SUV max ) was measured in the dominant intraprostatic lesions. These lesions were correlated with available clinical reports describing the prostatic lesions. SUV max was measured according to the E-PSMA criteria and was compliant with the EARL standards [10,20]. Volumes of interest (VOI) were manually drawn of the size of the prostate area, carefully omitting physiological activity from the bladder or urethra. SUV max was automatically calculated in those VOI using available clinical software: Sectra IDS7 v22.1 (Sectra AB, Linköping, Sweden) and Osirix v12.0 MD (Pixmeo SARL, Bernex, Switzerland). To cross-validate both software programs, identical scans from 10 patients were analysed using both programs, with 100% agreement.
Additionally, the tumour-to-blood ratio (TBR) was calculated in all patients since image-based TBR was found to better characterize the tumour than SUV max can in the quantification of [ 18 F]DCFPyL PET/CT scans [22]. The TBR was derived by normalizing the mean tumour uptake (Bq/cc) with the arterial blood activity concentration (Bq/cc) in the ascending aorta.

Statistical Analysis
Continuous variables were assessed for normality using histogram analysis and were summarized with median values and the inter-quartile range (IQR). Categorical variables were summarized with frequencies and proportions. Variables with non-normal deviation were log transformed for further analyses. To compare the medians of the SUV max with preoperative and postoperative variables, the Mann-Whitney-Wilcoxon test was used for two groups, and the Kruskal-Wallis test was used for three groups or more. Correlations between non-parametric continuous variables were assessed using Spearman's Rho. Cox regression was performed to identify variables that were associated with BCR-free survival. The follow-up time in Cox regression analysis was defined as the duration in days from RARP to the occurrence of BCR, or until the last recorded measurement of PSA level if the patient did not experience BCR. Correlated variables were analysed separately in survival analyses, as multi-collinearity reduces the precision of the estimated coefficients, which weakens the statistical power of the models. Kaplan-Meier plots were drawn to show the relationship between the SUV max and BCR in subgroups. For this purpose, the SUV max values were grouped and plotted as an independent value versus the percentage of patients with BCR in 1 and in 2 years as a dependent value. We visually assessed the risk of BCR per increased value of SUV max to determine low-risk and high-risk groups. The log-rank test was used to evaluate if those subgroups significantly differed from each other.
To analyse the predictive value of SUV max for BCR, proportional hazard models were created and compared. Differences and comparative suitability of models were assessed using both the likelihood ratio test and Akaike information criteria (AIC). Timedependent receiver operating characteristic (ROC) curves were generated to calculate the time-dependent area under the curve (AUC) for the prediction model of BCR. Harrell's C concordance indices and R-squared values were calculated for the two prediction models. Statistical significance was set at p < 0.05. Statistical analysis was performed using IBM SPSS Statistics for Windows ® V27 (Armonk, NY, USA: IBM Corp) and RStudio 4.2.1 for Windows ® . The figures were generated using GraphPad Prism ® software (Version 9.3.1 for Windows, GraphPad Software).

Baseline Characteristics
A total of 464 patients were included in this study. The baseline characteristics are presented in Table 1. All patients received PSMA PET/CT prior to RARP, of whom 238 (51%) patients were scanned with the tracer [ 68 (

Level of SUV max and Biochemical Recurrence (BCR) of Disease
Analysing the entire cohort, median SUV max was significantly higher in patients who experienced BCR after one and two years of follow-up compared to that of the BCR-free patients (12.02 vs. 7.69, p < 0.001 and 11.18 vs. 7.72, p = 0.002, respectively). This significant difference was also found via [ 68   The previously mentioned SUVmax low-and high-risk groups for BCR were visually assessed, characterising an SUVmax ≤ 10 as low risk and an SUVmax > 10 as high risk for developing BCR ( Figure S1). The analysis of all subgroups in the included cohort demonstrated that SUVmax > 10 was a significant predictor for BCR (p < 0.001) (Figure 2). SUVmax > 10 also showed to be a significant predicting factor for BCR in subgroups with bISUP 3-5, mT3 and an EAU classification high risk for BCR (p < 0.001, p = 0.002, p = 0.01, p = 0.004, respectively), but not in patients with mT2 (p = 0.07), or patients with EAU intermediate risk for BCR (p = 0.06). SUVmax > 10 was associated with BCR in univariate cox regression analysis as well (HR 1.87 (95%-CI 1.34-2.60), p= <0.001). The previously mentioned SUV max low-and high-risk groups for BCR were visually assessed, characterising an SUV max ≤ 10 as low risk and an SUV max > 10 as high risk for developing BCR ( Figure S1). The analysis of all subgroups in the included cohort demonstrated that SUV max > 10 was a significant predictor for BCR (p < 0.001) (Figure 2). SUV max > 10 also showed to be a significant predicting factor for BCR in subgroups with bISUP 3-5, mT3 and an EAU classification high risk for BCR (p < 0.001, p = 0.002, p = 0.01, p = 0.004, respectively), but not in patients with mT2 (p = 0.07), or patients with EAU intermediate risk for BCR (p = 0.06). SUV max > 10 was associated with BCR in univariate cox regression analysis as well (HR 1.87 (95%-CI 1.34-2.60), p ≤ 0.001). Data are presented in graphs comparing the maximum standardised uptake values (SUVmax) of the dominant intraprostatic lesion from below 10 to a value of 10 or higher in both the whole cohort as well as subgroups of the preoperative variables.

SUVmax in a Prediction Model for Biochemical Recurrence (BCR) of Disease
To analyse the added predictive value of the SUVmax, a baseline predictive model was created, incorporating the preoperative variables, log2PSA, bISUP, mT and the PI-RADS score. This model was subsequently compared with a new prediction model, wherein log2SUVmax was added to the variables of the baseline model. The inclusion of SUVmax improved the AIC value of the prediction model (1502 compared to 1498) and was significantly different from the baseline model (p = 0.02). After the addition of the SUVmax to the baseline model, higher time-dependent AUC values were found at every time point compared to those of the baseline model, as presented in Table 5. Additionally, the Harrel C concordance index was calculated for the baseline prediction model, as well as the new model, yielding nearly similar results of 68.2% and 69.7%, respectively. Similarly, the Rsquared values of the two models showed marginal improvements: 11.9% vs. 13.0%. Data are presented in graphs comparing the maximum standardised uptake values (SUV max ) of the dominant intraprostatic lesion from below 10 to a value of 10 or higher in both the whole cohort as well as subgroups of the preoperative variables.

SUV max in a Prediction Model for Biochemical Recurrence (BCR) of Disease
To analyse the added predictive value of the SUV max , a baseline predictive model was created, incorporating the preoperative variables, log 2 PSA, bISUP, mT and the PI-RADS score. This model was subsequently compared with a new prediction model, wherein log 2 SUV max was added to the variables of the baseline model. The inclusion of SUV max improved the AIC value of the prediction model (1502 compared to 1498) and was significantly different from the baseline model (p = 0.02). After the addition of the SUV max to the baseline model, higher time-dependent AUC values were found at every time point compared to those of the baseline model, as presented in Table 5. Additionally, the Harrel C concordance index was calculated for the baseline prediction model, as well as the new model, yielding nearly similar results of 68.2% and 69.7%, respectively. Similarly, the R-squared values of the two models showed marginal improvements: 11.9% vs. 13.0%.

Discussion
The aim of this study was to investigate whether the PSMA expression of the dominant intraprostatic lesion, defined as SUV max via primary staging PSMA PET/CT, was associated with BCR in patients with primary PCa prior to RARP. Two commonly used radiolabelled PSMA tracers were analysed: [ 68 Ga]Ga-PSMA-11 and [ 18 F]DCFPyL. We found that SUV max was an independent significant predictor for BCR in patients who undergone [ 68 Ga]Ga-PSMA-11 or [ 18 F]DCFPyL-120 PET/CT. In addition, SUV max > 10 was shown to be significantly associated with BCR in the entire patient cohort and for individual subgroups: bISUP 3-5, mT3 and patients with the EAU classification of a high risk for BCR.
Due to differences in the scan protocols, and thereby, differences in PSMA tracer uptake, the [ 18 F]DCFPyL PET/CT scans were divided into two groups based on the tracer incubation time: 60 min ([ 68 Ga]Ga-PSMA-11 and [ 18 F]DCFPyL-60) or 120 min ([ 18 F]DCFPyL-120) and were evaluated separately. Via univariable analysis, the Log 2 SUV max of the dominant intraprostatic lesion of the entire cohort as well as the Log 2 SUV max of the individual tracer subgroups were shown to be significant predictors for BCR. More importantly, via multivariate analysis, to assess multiple preoperative variables for their ability to predict BCR after surgery, the Log 2 SUV max was shown to be an independent significant predictor for the development of BCR in [ 68 Ga]Ga-PSMA-11 and [ 18 F]DCFPyL-120 PET/CT scans.
To our knowledge, this study is the first report to describe the association between SUV max and BCR for both [ 68 68 Ga]Ga-PSMA-11 PET/CT prior to RARP and showed a 5.5 fold increase in the hazard for BCR in patients with SUV max > 8 compared to SUV max < 8 via multivariate analysis [11].
Furthermore, a retrospective analysis of 186 patients with primary PCa who underwent [ 68 Ga]Ga-PSMA-11 PET/CT prior to surgery by Wang et al. [14] showed that patients with a localised disease and a higher SUVmax had a less favourable BCR-free survival rate (p = 0.02). In addition, a recent study by Roberts et al. [15] examining a larger cohort of 848 men who underwent [ 68 Ga]Ga-PSMA-11 PET/CT scans demonstrated this outcome as well (p < 0.001). The SUV max in [ 68 Ga]Ga-PSMA-11 PET/CT scans was, therefore, in line with our results, which indicate that it is an independent predictor for BCR. It is important to underline that the independent added value of SUV max for the prediction of BCR is often explained by the strong correlation between the SUV max and tumour grade [11,12,25], resulting in a higher risk for developing BCR.
A possible explanation for the non-significant correlations and associations with BCR in the [ 18 F]DCFPyL-60 group might be related to differences in tracer incubation compared to that of the [ 18 F]DCFPyL-120 patients [19]. Previous studies demonstrated a significant rise in tumour tracer uptake between 60 and 120 min after the tracer injection, leading to a higher tumour detection rate in [ 18 F]DCFPyL-120 patients [19,22]. Therefore, SUV max might not be suitable for universal application across different scan protocols with different tracer uptake intervals, without considering a translation factor and tracer decay.
Another notable difference in the scan protocols was the administered tracer dose in the two [ 18 F]DCFPyL scan groups. In the [ 18 F]DCFPyL-120 group, a median tracer dose of 312 MBq was injected, compared to 200 MBq in patients who were scanned after 60 min. However, the injected tracer doses are accounted for in the SUV max calculation; therefore, we expect that those differences in tracer dosage have a negligible impact on intergroup SUV max variability. Moreover, different PET/CT scanners were used in this study, with all [ 18 F]DCFPyL-120 PET/CT scans performed on the same PET/CT scanner, while [ 18 F]DCFPyL-60 and [ 68 Ga]Ga-PSMA-11 PET/CT scans were performed interchangeably using two different scanners. The use of different scanners may introduce some variability in the SUV max values; although, this variability is limited for EARL-accredited scanners [26]. Further investigation is required to determine if the SUV max can be directly applied to various scan protocols, PET/CT scanners, or if the implementation of a correction factor is necessary.
Previous research by Jansen et al. suggested that studying the TBR is a superior simplified method for determining the PSMA tracer intensity expression in the SUV max in [ 18 F]DCFPyL PET/CT scans [22]. Therefore, we repeated our multivariate analysis, including the TBR instead of the SUV max . This yielded an even more significant association with BCR for [ 18 F]DCFPyL-120 and a slightly negative, non-significant association for [ 18 F]DCFPyL-60. These outcomes align with previous pharmacokinetic results, showing that the TBR is highly influenced by the interval of tracer uptake. We therefore underline the importance of scanning patients 120 min after an injection when using the [ 18 F]DCFPyL tracer.
Another possible explanation for the differences between the scan groups might be the smaller sample size of the two individual [ 18 F]DCFPyL scan groups compared to that of the [ 68 Ga]Ga-PSMA-11 group, as no major differences in surgical demographics between the groups were present.
In this study, we observed that patients with high SUV max values (SUV max > 10) and less-favourable preoperative variables had significantly higher BCR rates compared to those of the patients with low SUV max values (SUV max ≤ 10) in the same subgroups. The identification of factors that could aid in predicting survival outcomes is crucial, since postoperative disease recurrence is challenging to predict [5]. Therefore, we believe that incorporating the SUV max /TBR of primary PET/CT scans into preoperative prediction models could be of added value.
Furthermore, we found that the prediction model incorporating both preoperative variables and SUV max is superior to the baseline prediction model that only included preoperative variables. As a result, slightly higher values for the area under the curve (AUC) were observed at various time points. Also, slightly higher R-squared values and C-indexes were found. Notably, a significant improvement of the AIC was found when SUV max was added to the baseline prediction model. These findings are in line with literature, as PSMA PET/CT is increasingly being incorporated into preoperative risk stratification models. Qui et al. [27] retrospectively investigated the prognostic role of preoperative [ 68 Ga]Ga-PSMA-11 PET/CT scans in predicting BCR after RARP. They evaluated 77 patients after 2 years of follow-up and demonstrated the superior discriminative ability of a risk model incorporating SUV max compared to that of the CAPRA and D'Amico models. The exact additional value of SUV max in these models needs to be explored further in future, larger cohort studies.
A strength of our study was that VOI's delineation of the dominant intraprostatic lesions was performed by one expert. Therefore, no inter-observer variability occurred. Moreover, delineating VOI, and thus, calculating the SUV max is a simple action and can, therefore, easily be implemented in current clinical practice. A disadvantage of this method is that manual VOI alteration is necessary when the suspected dominant intraprostatic lesion is close to the bladder wall to rule out bladder activity interference. To solve this issue, Artificial intelligence tools might play an important role in the future.
This study has several limitations that need to be acknowledged. Firstly, it was conducted in a retrospective setting, which introduces the possibility of selection bias. Secondly, we did not incorporate the total tumour volume (TTV) as a variable in this study. Considering that the SUV max is also influenced by tumour volume, the inclusion of the TTV would have provided more accurate SUV max values [22], since the TTV also has prognostic abilities [24]. Instead, we demonstrated that utilizing the PI-RADS score as a surrogate for tumour volume, wherein a tumour volume of ≥15mm via MRI differentiates between a PI-RADS 4 or 5 lesion [28], yielded satisfactory results. Thirdly, the bi-centric setup resulted in different scan protocols, thereby leading to variations in the obtained results, as previously stated.
Based on the present findings, SUV max could contribute to the prediction of BCR in patients with primary PCa before undergoing RARP. Future, prospective studies are required to re-evaluate whether the SUV max of the dominant intraprostatic lesion can be considered as an addition to prediction models. Moreover, larger studies are needed to evaluate the translatability of SUV max to different scan protocols.

Conclusions
This study evaluated the prognostic value of SUV max for the biochemical recurrence of disease after RARP in patients with PCa undergoing [ 68 Ga]Ga-PSMA-11 and [ 18 F]DCFPyL PET/CT. SUV max proved to be an independent predictive factor for BCR in patients with primary PCa and in patients who underwent [ 68 Ga]Ga-PSMA-11 PET/CT or [ 18 F]DCFPyL PET/CT and were scanned 120 min after tracer injection. In these tracer groups, the median SUV max was significantly higher in patients with BCR after 1 year of follow-up. Moreover, SUV max > 10 was a significant predictor for BCR. Therefore, the SUV max in PSMA PET/CT may be of added value to risk stratification models for BCR in patients with primary PCa.
Supplementary Materials: The following supporting information can be downloaded at: https://www. mdpi.com/article/10.3390/diagnostics13142343/s1, Figure S1: Histograms presenting the percentage of patients who experienced biochemical recurrence (BCR) or remained BCR-free divided in groups by maximum standardised uptake values (SUV max ).