PSMA-ligand uptake can serve as a novel biomarker in primary prostate cancer to predict outcome after radical prostatectomy

Background The prostate-specific membrane antigen (PSMA) is a relevant target in prostate cancer, and immunohistochemistry studies showed associations with outcome. PSMA-ligand positron emission tomography (PET) is increasingly used for primary prostate cancer staging, and the molecular imaging TNM classification (miTNM) standardizes its reporting. We aimed to investigate the potential of PET-imaging to serve as a noninvasive imaging biomarker to predict disease outcome in primary prostate cancer after radical prostatectomy (RP). Methods In this retrospective analysis, 186 primary prostate cancer patients treated with RP who had undergone a 68Ga-PSMA-11 PET up to three months prior to the surgery were included. Maximum standardized uptake value (SUVmax), SUVmean, tumor volume (TV) and total lesion (TL) were collected from PET-imaging. Moreover, clinicopathological information, including age, serum prostate-specific antigen (PSA) level, and pathological characteristics, was assessed for disease outcome prediction. A stage group system for PET-imaging findings based on the miTNM framework was developed. Results At a median follow-up after RP of 38 months (interquartile range (IQR) 22–53), biochemical recurrence (BCR) was observed in 58 patients during the follow-up period. A significant association between a positive surgical margin and miN status (miN1 vs. miN0, odds ratio (OR): 5.428, p = 0.004) was detected. miT status (miT ≥ 3a vs. miT < 3, OR: 2.696, p = 0.003) was identified as an independent predictor for Gleason score (GS) ≥ 8. Multivariate Cox regression analysis indicated that PSA level (hazard ratio (HR): 1.024, p = 0.014), advanced GS (GS ≥ 8 vs. GS < 8, HR: 3.253, p < 0.001) and miT status (miT ≥ 3a vs. miT < 3, HR: 1.941, p = 0.035) were independent predictors for BCR. For stage I disease as determined by PET-imaging, a shorter BCR-free survival was observed in the patients with higher SUVmax (IA vs. IB stage, log-rank, p = 0.022). Conclusion Preoperative miTNM classification from 68Ga-PSMA-11 PET correlates with postoperative GS, surgical margin status and time to BCR. The association between miTNM staging and outcome proposes 68Ga-PSMA-11 PET as a novel non-invasive imaging biomarker and potentially serves for ancillary pre-treatment stratification. Supplementary Information The online version contains supplementary material available at 10.1186/s13550-021-00818-2.

widely used for biochemical recurrence (BCR) prediction. However, there is growing interest in identifying novel biomarkers to improve BCR prediction accuracy of prostate cancer patients after radical prostatectomy (RP).
In the last few years positron emission tomography (PET) probes targeting prostate-specific membrane antigen (PSMA) has significantly improved detection and localization of disease in primary and recurrent prostate cancer [1, [10][11][12]. PSMA is a type II integral membrane glycoprotein with folate hydrolase activity, internalization after activation and is encoded by the FOLH1 gene [13,14]. PSMA expression increases progressively in highergrade prostate tumor cells and metastatic lesions [15,16].
Increased PSMA expression in immunohistochemistry was more often observed in pathological stage III or IV tumors (51%) compared to stage I and II tumors (32%, p = 0.029) [17]. High-level PSMA expression in immunohistochemistry was associated with a higher risk of BCR and overall survival in several studies [16,17]. Finally, expression of membranous PSMA is also associated with higher rates of defective deoxyribonucleic acid (DNA) damage repair gene [18].
In the last two decades, for different tumor entities results from imaging were introduced as non-invasive quantitative biomarkers. For fluorodeoxyglucose (FDG)-PET-imaging, Wieder et al. have demonstrated that mean standardized uptake value (SUV mean ) can be used to preoperatively predict histopathological response in esophageal squamous cell carcinoma (ESCC) patients. A decrease in SUV mean of 44% ± 15% from responders and 21% ± 14% from non-responders (p = 0.0055) was observed after radiochemotherapy [19]. Besides, Giovacchini et al. have observed that the positive results of 11 C-choline PET/computed tomography (CT) predict prostate cancer-specific survival in patients after RP [20]. For metastatic castration-resistant prostate cancer (mCRPC), the bone scan index (BSI) [21] and quantitative parameter from PET have been reported to serve as potential predictive biomarkers for bone tumor burden [22,23].
To standardize reporting of PSMA-targeted PETimaging, a unified molecular-imaging TNM classification (miTNM, version 1.0) has been recently introduced [24]. It is envisioned that its system classifying tumor extent similar to the pathological TNM-system might serve as qualitative imaging biomarker potentially stratifying disease outcome.
The aim of our retrospective analysis was to investigate the potential of quantitative and qualitative parameters from 68 Ga-PSMA-11 PET to serve as non-invasive imaging biomarkers to predict BCR in primary prostate cancer after RP allowing for ancillary preoperative risk stratification.

Patient selection
We screened the institutions´ database for all patients who underwent 68 Ga-PSMA-11 PET imaging maximum 3 months prior the RP between January 2013 and August 2017. All patients who received neoadjuvant therapy prior to RP had PSA-persistence after RP or in whom follow-up data were missing were excluded. Finally, 186 patients with D'Amico intermediate-to high-risk primary prostate cancer were included in this retrospective study (Additional file 1: Fig. S1). Table 1 summarizes the clinical and histopathological characteristics. BCR was defined as a serum PSA level rising above 0.2 ng/ml. The primary endpoint was time to BCR. The time to BCR was calculated from the date of surgery. The retrospective analysis has been approved by the Ethics Committee of the Technical University Munich (750/20 S-KH).

Imaging protocol
The synthesis of 68 Ga-PSMA-11 [25] was described previously [26]. Patients were intravenously injected with a median of 139 MBq of 68 Ga-PSMA-11 (interquartile range (IQR) 112-156). PET acquisition was started at a median of 54 min (IQR 49-65) after the tracer injection. Nighty-three patients underwent 68 Ga-PSMA-11 PET/CT using a Biograph mCT flow scanner (Siemens Medical Solutions, Erlangen, Germany), and 93 patients underwent 68 Ga-PSMA-11 PET/magnetic resonance imaging (MRI) using an integrated whole-body PET/MRI system (Siemens Biograph mMR, Erlangen, Germany). Details on PET/CT and PET/MRI acquisition were described previously [27,28]. 68 Ga-PSMA-11 PET/CT and 68 Ga-PSMA-11 PET/ MRI images were evaluated by one nuclear medicine physician blinded to the postoperative histopathological results. All lesions were reannotated by two experienced board-certified nuclear medicine physicians. Any focal or diffuse tracer uptake in the prostate or extra-prostatic lesions above the surrounding background and not associated with physiological uptake was considered suspicious for malignancy. One circular region in transaxial slices was drawn over the prostate and over every extra-prostatic lesion automatically adapted to a three-dimensional volume of interest (VOI) using Syngo.Via (Siemens Healthineers, Erlangen, Germany). A 40% isocontour of the SUVmax was used to determine the SUVmean. Typical pitfalls in PSMA-ligand PET including low to moderate PSMA uptake correlated with osteoblastic changes (i.e., fractures or degenerative changes), celiac, and ganglia were taken into consideration [29]. SUV max , SUV mean , tumor volume (TV), and total lesion (TL) of every VOI were calculated. Prostatic and extra-prostatic lesions were classified according to miTNM classification [24]. In PSMA-PET negative primary tumors, the CT or MRI part of the hybrid PET-exam was used for miTNM stage group determination. Similar to the structure of the American Joint Committee on Cancer (AJCC) Prostate Cancer Prognostic Stage Groups [30], we established a stage group system using the different grades from the miTNM staging system (Table 2). To allow the discrimination into different risk groups based on the characteristics of the primary tumor, we used a SUV max cut-off of 5.4 to subgroup stage I disease into stage IA and IB. The cut-off was derived from a recent study proposing it the optimal cut-off to distinguish between GS ≤ 7a and GS ≥ 7b [31].

Statistical analysis
Descriptive statistics were used to display continuous variables as the median and IQR with 25th and 75th percentiles (Q1-Q3), mean ± standard deviation (SD), as well as percentages. The association between pathological results and 68 Ga-PSMA-11 PET findings was investigated with uni-and multivariate Logic regression analyses, and the corresponding odds ratios (OR) and 95% confidence intervals (CI) were calculated. Postoperative BCR-free survival was estimated using the Kaplan-Meier method and compared between groups using the Log-rank test. Moreover, uni-and multivariable Cox regression analysis were performed to determine the ability of clinicopathological factors and 68 Ga-PSMA-11 PET findings to predict BRC after RP, and the corresponding hazard ratios (HR) and 95% CI were calculated. The multivariable model only included parameters with a significant association on univariable analysis. A p value of 0.05 was used as the cut-off for statistical significance.
Given its low sample size (n = 7) the miM1 subgroup was excluded for univariable and multivariable analysis.
Statistical evaluation was performed with IBM SPSS Statistics Version 20 (Armonk, NY, USA), and the figures were generated using GraphPad Prism Version 8 (San Diego, California, USA).

Discussion
Our retrospective analysis demonstrates the potential predictive value of 68 Ga-PSMA-11 PET findings for BCR-free survival of primary prostate cancer after RP. Information from the recently introduced miTNM classification is associated with outcome. In addition, several correlations between quantitative and qualitative measures from 68 Ga-PSMA-11 PET and pathological parameters have been observed. Prognostic tools of BCR are required and essential for improving treatment management of prostate cancer patients and reducing prostate cancer-associated mortality of patients developing BCR after primary treatment [32]. With the successful application of PSMA-ligand PET for primary staging in prostate cancer patients, clinical studies are necessary to investigate its predictive value. PSMA-ligand PET is increasingly used for selection, monitoring and individualization of prostate cancer treatments. The present analysis is the first to investigate the association of miTNM classification from preoperative 68 Ga-PSMA-11 PET imaging and postoperative histopathological findings and the potential of miTNM reporting to serve as predictors for BCR after RP. Consequently, we performed a prognostic validation of the miTNM system as a framework for PSMA-ligand PET reporting in a relatively large patient cohort. Preoperative 68 Ga-PSMA-11 PET and miTNM classification could help to stratify risk for BCR after RP and could potentially further influence the clinical patient management.
In addition, and novel compared to the current literature our data indicate that also parameters from 68 Ga-PSMA-11 PET might serve as non-invasive biomarkers. We identified a miT stage ≥ 3a in 68 Ga-PSMA-11 PET as surrogate for higher GS (OR: 2.696, 95% CI 1.39-5.23, p = 0.003) and worse BCR-free survival (HR: 1.941, 95% CI 1.047-3.599, p = 0.035). Notably, pelvic lymph node metastases in 68 Ga-PSMA-11 PET were not detected as an independent predictor for BCR in this study (HR: 1.233, 95% CI 0.389-3.908, p = 0.722). However, the BCR-free survival differed significantly between miN0 and miN1 group (log-rank, p = 0.005). Interestingly also Raheem et al. have failed to detect lymph nodes in histopathology as an independent predictor to BCR after RP in a study including 359 patients [4]. Of note, the sample size of miN1 group (n = 18) in our cohort was relatively small. Thus, the interpretation of results should be with caution, and further studies including more miN1 patients are needed to clarify the predictive value of miN classification for BCR after RP. Besides, our results indicate a negative association of TV with BCR-free survival (HR: 0.934, 95% CI 0.883-0.988, p = 0.017). Contrarily, Choi et al. have reported a significantly higher BCR-free survival rate in pT2 prostate cancer patients with percent tumor volume ≤ 7.5%, which was assessed using histological samples (p < 0.001) [33]. This is partly related to the methods of obtaining tumor volume and studies are necessary to assess the standard of TV calculation from 68 Ga-PSMA-11 PET and pathological samples.
With this work we also introduced a stage grading system based on the recent proposed molecular staging system (miTNM staging system, version 1.0) combined with quantitative parameters. It is intended to mirror the AJCC staging system based on clinicopathological parameters which has proven to be a fundamental tool that also informs treatment decisions [34]. Bhindi et al. have confirmed the ability of the 8 th edition to predict oncologic outcomes [35]. However, the AJCC staging system utilized clinical or pathological TNM stage and no parameters from imaging. With the increasing use of PSMA-ligand PET in clinical routine, a logical next step is to use information from non-invasive imaging prior to definite treatment for risk stratification.
We have shown that the miT stage is an independent predictor of BCR, and we observed a widely Fig. 3 Kaplan-Meier curves comparing biochemical recurrence-free survival of selected patients stratified by a tumor volume, b total lesion, c SUV mean and d SUV max . Longer biochemical recurrence-free survival was associated with lower SUV mean and SUV max varying prognosis in the miT2 stage patients. Similarly, a recent study has revealed that high intraprostatic 68 Ga-PSMA-11 uptake (SUV max > 8) predicts short progression-free survival rate among patients with GS 3 + 4 on biopsy [36]. The significant difference in BCR-free survival rate has been confirmed in IA and IB stage groups. Our findings propose that a SUV max cut-off extracted from literature could further stratify the group of miT2 Table 6 Univariable analysis for the association of baseline factors with BCR-free survival BCR biochemical recurrence, CI confidence interval, iPSA initial PSA, IQR interquartile range, LN lymph node, PET positron emission tomography, PSA prostate-specific antigen, PSMA prostate-specific membrane antigen, SUV standardized uptake value, TL total lesion, TV tumor volume *Significant associations are given in bold primary disease into patients with more aggressive disease and worse prognosis. Further studies are necessary for prognostic validation of other stage groups. The present study has several limitations. It is a retrospective analysis and includes only patients from a single center, which can introduce potential bias. Despite inclusions of a large number of patients, the sample size of patients in the miM1 group was too small to conduct meaningful analysis. This is mainly related to the fact that most patients with extrapelvic metastases do not undergo primary curative RP but either get systemic treatment with or without local treatment. This explains the low number of patients in miTNM stage group III and IV. In summary, further prospective investigations with large patient numbers are necessary to fully investigate the potential of the miTNM staging and our proposed grading system to predict patient outcome after curative intent RP.

Conclusion
Our retrospective analysis indicates that the miTNM framework developed to standardize PSMA-ligand PET reported is independently associated with BCRfree survival of primary prostate cancer after RP. We demonstrated significant associations between 68 Ga-PSMA-11 PET findings and histopathological parameters. In summary, our results outline that the miTNM classification and the presented further development of a Table 7 Multivariable analysis for the association of baseline factors with BCR-free survival BCR biochemical recurrence, CI confidence interval, iPSA initial PSA, LN lymph node, PET positron emission tomography, PSA prostate-specific antigen, PSMA prostatespecific membrane antigen, SUV standardized uptake value, TV tumor volume *Significant associations are given in bold miTNM-based stage group system can serve as non-invasive imaging biomarkers of risk stratification for primary prostate cancer patients. However, further and prospective studies including patients with different treatments and stages are needed to fully assess the predictive value of PSMA-ligand PET imaging in the setting of newly diagnosed prostate cancer.