Introduction

Prostate cancer (PCa) is the most common cancer in men in the Western world [1, 2]. Histopathological verification is required to confirm the diagnosis and is standardly attained through ultra-sound-guided systematic prostate biopsies [3]. These random biopsies are subject to sampling error, however, resulting in false-negative outcomes and imprecise tumour-risk assessment [4, 5]. To overcome this, multi-parametric magnetic resonance imaging (mpMRI) prior to systematic biopsy has been implemented in clinical guidelines, enabling targeted biopsies of radiologically suspected lesions (MRI-TBx) [3, 6,7,8].

Besides conventional imaging modalities such as mpMRI, novel imaging techniques including prostate-specific membrane antigen-positron emission tomography/computed tomography (PSMA-PET/CT) have been introduced. PSMA is significantly overexpressed in malignant prostate cells, correlates with higher tumour grades and represents a marker of tumour aggressiveness [9, 10]. PSMA-PET/CT-imaging has been shown to accurately identify the primary prostate tumours, with detection rates of 98–100% [11, 12]. PSMA-PET/CT could thus be used to localize and guide targeted prostate biopsy in patients with clinically suspected PCa. Furthermore, PSMA-PET/CT would simultaneously provide screening for bone and lymph-node metastases, as it is repeatedly found to be more sensitive than conventional imaging (i.e., MRI, bone scintigraphy and CT) in the initial staging setting [13, 14].

This is the first prospective study on the accuracy of 18F-DCFPyL (PSMA) PET/CT imaging for the primary detection of PCa. The primary aim was to assess the accuracy of 18F-DCFPyL-PET/CT to localise primary PCa within the prostate gland, by comparing imaging results from 18F-DCFPyL-PET/CT to final histopathology of the robot-assisted radical prostatectomy (RARP) specimen. The secondary objectives were to investigate the ability of 18F-DCFPyL-PET/CT to provide a recommendation for potential targeted biopsy and to assess the diagnostic accuracy of determining local tumour stage (pT).

Methods

Study design and patient population

This was a prospective, non-randomised study in patients with diagnosed primary PCa. Pre-operative imaging results were compared to histopathology following RARP. All subjects signed informed-consent for the collection of their clinical data. The study has been approved by the ethical review board of the Amsterdam University Medical Centre (AUMC) (review number 2017.543). Patients were enrolled consecutively between May 2018-May 2019 in Amsterdam UMC, location VUmc.

Patients had histologically proven, intermediate or high-risk, PCa, for which they underwent RARP [3, 15]. Of all included patients, age, prostate volume, initial prostate-specific antigen (PSA)-level, clinical T-stage, pathological biopsy features (histopathological grade, number of cores with cancer) and European Association of Urology (EAU)-risk category were collected [3, 15]. A 12-segment anatomic mapping model of the prostate was used to localise and characterise the prostatic tumours, with 2 additional segments representing the seminal vesicles (pT3b) (Appendix-1 in ESM) [16].

Imaging protocol

Patients were staged with 18F-DCFPyL which was synthesised under Good Manufacturing Practices conditions, as described by Jansen et al. [17, 18]. PET images were acquired at a median of 118 min after injection of the radiotracer (interquartile range [IQR] 113–122 min) with a median dose of 313 MBq 18F-DCFPyL (IQR 299–324 MBq), and a median of 5.4 weeks (IQR 3.0–7.2) prior to surgery. Image-acquisitions were performed using a Philips Ingenuity TF (Philips Healthcare®, NL/USA)-PET/CT system. No diuretics were administered prior to the scan. The scan trajectory included mid-thighs to skull-base, with 4 min per bed position. All PET scans were combined with a diagnostic CT scan (110 mAs, 120 kV), without contrast-enhancement. Images were corrected for decay, scatter, random coincidences, and photon attenuation.

Images were reconstructed with a BLOB-based Ordered-Subsets Expectations Maximization algorithm (Philips, 3 iterations; 33 subsets) [19]). The default Ordered-Subsets Expectations Maximization with Time-of-Flight reconstruction was used. Reconstructions included both 4 mm for semi-quantification purposes, and 2 mm slices for visual interpretation (matrix size 144 × 144, slice thickness 4 mm; matrix size 288 × 288, slice thickness 2 mm).

Image interpretation and 18F-DCFPyL-based potential biopsy recommendation

Scan interpretation was performed blinded for the pathology results and other imaging by two nuclear medicine physicians (DO,GZ) with ample experience in 18F-DCFPyL-PET/CT reading (> 300 scans), in consensus. The readers used the 12-segment mapping model to demarcate the image-detected tumour extent (Appendix-1 in ESM) [20]. For all positive segments, the readers’ diagnostic confidence was evaluated using a five-point scale, alike the PSMA-RADS classification [21] (score 1–2 ‘benign’; 3 ‘equivocal’; 4–5 ‘likely PCa’). PSMA-RADS 4–5 were defined as suspicious for PCa (‘positive’), and were used for the final diagnostic accuracy analysis. Based on PET/CT-imaging, two segments per patient were indicated to be potentially targeted by prostate biopsy. These segments were selected based on both visual interpretation (location, size) and semi-quantification by determining the highest standardised uptake values (SUVmax) of the suspected lesions. Finally, the readers indicated if radiological extra-capsular extension (ECE, rT-stage 3a) or invasion into the seminal vesicles (rT3b) was suspected.

Pathology analysis

RARP specimens were processed according to clinical routine and the International Society of Urological Pathology (ISUP) guidelines [22]. All specimens were fixated in formaldehyde (10%) directly after surgery. The surface of the specimens was inked and the apex and base (bladder neck) were removed. The mid part of the specimen was cut perpendicular to the urethra in 4 mm slices. The apex and base parts were cut in a sagittal fashion. Histologic slices were produced after sectioning in quadrants. Blinded by PET/CT results, an experienced uro-pathologist (PV) reviewed all slices and delineated all tumour depositions on the 12-segment mapping model of the prostate and the 2 segments of the seminal vesicles (Appendix-1 in ESM). For all segments that contained tumour, a Gleason score and ISUP grade was provided and the presence of ECE (pT3a). Clinically significant PCa (csPCa) was defined as PCa with Gleason score ≥ 3 + 4 = 7 (ISUP grade ≥ 2) or any tumour with ≥ pT3a. The index lesion was defined as the largest lesion with the highest ISUP grade or stage.

Statistical analysis

The localisation of the detected prostate tumour by 18F-DCFPyL-PET/CT was matched to the histopathology results and the sensitivity, specificity, positive predicting value (PPV), and negative predicting value (NPV) were calculated on a segment basis. Correlation of 18F-DCFPyL-PET/CT with histopathology was considered if exactly the same segment was demarcated (total agreement). Since there are no anatomical landmarks to delineate the segments, artificial segmentation can occur, causing a mismatch between the PET/CT- and histopathological findings while both correspond with the same lesion. Therefore, a second analysis of diagnostic accuracy was performed, in which PET correlation was also considered if there was a discrepancy of up to 1 region in the coronal or sagittal plane (near-total agreement) [23, 24]. Reciever‐operating characteristic curves (ROC) and area under the curve (AUC) analysis were performed to explore the accuracy of PSMA-PET/CT in the detection of segments containing csPCa based on the 5-point scale. For the assessment of pathological tumour stage (pT), we investigated the accuracy of 18F-DCFPyL-PET/CT to differentiate locally advanced disease (> rT3a) from prostate-confined disease (rT2). Numerical variables were summarised with median values and interquartile ranges (IQR); categorical variables with proportions (%). To compare medians of non-parametric data, the Mann–Whitney-Wilcoxon test and the Kruskal-Wallis test were used (significance set at p < 0.05). Statistical analysis was performed with IBM® SPSS® Statistics for Windows®, version 26.

Results

Patient characteristics

A total of 30 patients was included in this study, having a median initial PSA-level of 11.1 ng/ml (IQR 5.8–22.4). According to EAU guidelines, 10/30 (33.3%) patients had intermediate-risk PCa and 20/30 (66.6%) had high-risk PCa [15]. Pre-operative and post-operative characteristics of included patients are listed in Table 1.

Table 1 Pre- and postoperative characteristics of patients undergoing 18F-DCFPyL-PET/CT before robot-assisted radical prostatectomy
Full size table

Accuracy of 18F-DCFPyL-PET/CT to detect local prostate cancer on a segmental level

All patients showed PSMA expression in the prostate. In 30 evaluated patients, 420 segments (12 prostate segments + 2 seminal vesicle segments per patient) could be used both for PET/CT and histopathological mapping evaluation. PCa was present in 129 of the 420 (30.7%) segments on histopathological examination, and csPCa was found in 122 of the 420 segments (29.0%) (median 3 segments per patient, IQR 2–5). The sensitivity, specificity, PPV and NPV of 18F-DCFPyL-PET/CT to detect csPCa per segment with total agreement was 61.4% (95%CI 52.2–70.0%), 88.3% (95%CI 83.9–91.6%), 68.1% (95%CI 58.5–76.6%), and 84.8% (95%CI 80.2–88.5%), respectively (Appendix-2 in ESM). For near-total agreement, the sensitivity, specificity, PPV and NPV of 18F-DCFPyL PET/CT to detect csPCa per segment was 84.4% (95%CI 76.5–90.1%), 97.0% (95%CI 94.1–98.5%), 92.0% (95%CI 84.9–96.0%), and 93.8% (95%CI 90.3–96.1%), respectively. The area under the curve (AUC) of 18F-DCFPyL-PET/CT was 0.78 (95%CI 0.73–0.84) for the total agreement scores, and 0.85 (95%CI 0.80–0.90) for the near-total agreement scores (Appendix-3 in ESM). True positive segments had a median SUVmax of 8.26 (IQR 5.25–11.40), which was significantly higher than the median SUVmax of false-positive segments of 4.06 (IQR 3.56–5.10) (p = 0.02). The median SUVmax of true positive segments did not correlate with ISUP grade groups (p = 0.95) (Appendix-4 in ESM).

Potential targeted biopsy recommendation on a patient level

The primary potential biopsy recommendation by the nuclear medicine physician harboured csPCa in 24/30 (80.0%) patients and detected the index PCa lesion in 23/30 (76.6%) patients. When both the primary and secondary recommended segments would potentially be targeted, 18F-DCFPyL-PET/CT revealed csPCa in 28/30 (93.3%) patients. Moreover, it pinned the index PCa lesion in 26/30 (86.7%) patients. An example of potential 18F-DCFPyL-guided biopsy recommendation and concurrent histopathological examination of the RARP specimen is shown in Fig. 1. Potential biopsy recommendation that matched the index PCa lesion had a median SUVmax 8.62 (IQR 6.41–12.62). The recommendation for potential biopsy from the nuclear physicians that matched csPCA had a median SUVmax of 8.55 (IQR 6.34–13.79), and was significantly higher than the recommended potential biopsy segments that did not contain csPCa (median SUVmax of 3.10 (IQR 2.86–3.87) (p = 0.02).

Fig. 1
figure 1

67-Year-old man with cT3a, Gleason score 3 + 4 = 7 prostate cancer and an initial PSA 27 ng/ml considered candidate for radical prostatectomy with extended pelvic lymph-node dissection. a Schematic drawing of the 12 prostate segments included in the study. b Transversal fused 18F-DCFPyL-PET and CT show intense focal uptake in the right posterior midgland and apex segments, with extracapsular extension. Based on the highest SUVmax of 6.09 and tumour size, segment 5 is recommended for potential targeted biopsy. c histopathology (hematoxylin and eosin stain, original magnification × 10) shows a tumour focus, in both segment 5 and 1 with a Gleason score 3 + 4 = 7 prostate cancer with extraprostatic extension (pT3a), hereby confirming the index lesion localisation by PET

Full size image

Local staging

Final histopathological analysis revealed pT3a in 10/30 (33.3%) patients, and pT3b in 4/30 (13.3%) patients. The sensitivity, specificity, PPV and NPV of 18F-DCFPyL-PET/CT to detect locally advanced tumour growth (≥ pT3a) was 35.7% (95%CI 14.0–64.3%), 93.8% (95%CI 67.7–99.7%), 83.3% (95%CI 36.5–99.1%), and 62.5% (95%CI 40.8–80.4%), respectively (Appendix-5 in ESM). For the detection of pT3a sub-stage, the sensitivity, specificity, PPV and NPV of 18F-DCFPyL-PET/CT were 20.0% (95%CI 3.5–55.8), 100% (95%CI 80.0–100), 100.0% (95%CI 19.7–100), and 71.4% (95%CI 51.1–86.0), respectively. For the detection of pT3b sub-stage, the sensitivity, specificity, PPV and NPV of 18F-DCFPyL-PET/CT were 75.0% (95%CI 21.9–98.7), 92.3% (95%CI 73.4–98.7), 60.0% (95%CI 17.0–92.7), and 96.0% (95%CI 77.7–99.8), respectively.

Discussion

This is the first prospective study in which 18F-DCFPyL-PET/CT imaging was used to locate primary PCa within the prostate gland, exploring the diagnostic potential of PSMA-based targeted biopsies. A total of 30 patients diagnosed with intermediate and high-risk PCa that underwent 18F-DCFPyL-PET/CT prior to RARP was analysed. When using a prostate-mapping model, the potential 18F-DCFPyL-PET/CT-based targeted biopsy recommendation detected csPCa in 28/30 (93.3%) patients. Moreover, it detected the index PCa lesion in 26/30 (86.7%) patients. Potentially, this makes PSMA-targeted biopsy a diagnostic tool that may adequately guide precision prostate biopsy. In biopsy-naive patients at increased risk of metastatic spread, and in whom staging imaging is mandatory (e.g. PSA ≥ 20), 18F-DCFPyL-PET/CT could potentially be used simultaneously to stage patients and to target PSMA-avid prostatic lesions suspicious for PCa.

18F-DCFPyL-PET/CT imaging demonstrated a moderate per segment-based sensitivity for the detection of csPCa of 61.4%, at a 88.3% specificity. The moderate sensitivity indicates that 18F-DCFPyL-PET/CT was not able to detect all localised csPCa. Segmentation of the prostate gland is problematic, as no clear anatomical landmarks are available to delineate the different segments within the prostate (Appendix-1 in ESM). A tumour located on the border of the apex and middle part of the prostate could be classified in different segments by the nuclear medicine physician and uro-pathologist, while in fact, they detected the same lesion. Therefore, the near-agreement score was introduced to approximate clinical reality. The sensitivity of the near-agreement score of 18F-DCFPyL-PET/CT imaging for the detection of csPCa was higher at 84.4% with a specificity of 97.0%.

MpMRI has found a prominent place in the identification and localisation of PCa [8]. Few studies have directly compared the outcome of mpMRI to that of PSMA-PET/CT for the detection of localised PCa. Scheltema et al. analysed 56 patients with intermediate-risk PCa who underwent 68Ga-PSMA-PET/CT and mpMRI prior to RARP [20]. This study used the same 12-segment-based model of the prostate, and a leniency method similar to our near-agreement score. The patient-based sensitivity for detecting ISUP grade 2–3 PCa was 100% for PSMA-PET/CT vs 98% for mpMRI (PI-RADS 3–5). The segment-based sensitivity and specificity for PSMA-PET/CT was 88% (95% CI 83–92) and 93% (95% CI 91–95) compared 68% (95% CI 61–75) and 91% (95% CI 87–93) for mpMRI (PI-RADS 3–5).

In another study, Kesch et al. [23] studied 10 patients with primary high-risk PCa who underwent 18F- PSMA-1007-PET/CT and mpMRI with subsequent RARP. Nine of the men were diagnosed with MRI-TBx and 1 with systematic biopsy. Using a 36-segment mapping model, a similar assessment of agreement and near-total agreement was used. In 10 patients, 18F-PSMA-1007 PET/CT showed a high sensitivity (71% for total and 93% for near-total agreement), specificity (81% for total and 92% for near-total agreement), and accuracy (75% for total and 93% for near-total agreement) for the detection of csPCa. Although the specificity was similar, this study did however show lower sensitivity for PSMA-PET/CT compared to mpMRI (86% for total and 92% for near-total agreement). Above mentioned studies implicate that PSMA-PET/CT imaging performs at least equal to mpMRI to locate primary PCa. The rates of mpMRI might have been overestimated as at least a part of included patients in previously mentioned studies were diagnosed by MRI-TBx. Thus, selection bias may have been introduced.

There is therapeutic importance to distinguish between T2 and T3 disease (i.e. for planning nerve-sparing surgery, to opt for active surveillance) [3]. A moderate sensitivity for the detection of pT3a-b of 35.7% was observed using 18F-DCFPyL-PET/CT. So, similar to mpMRI, a substantial number of patients with ≥ pT3a was understaged by PSMA-based imaging [8]. However, in a majority of patients, a PSMA-PET/CT rT3a-b finding was confirmed after histopathological examination (PPV for pT3a-b disease of 83.3%). Moreover, the promising specificity for pT3a-b using 18F-DCFPyL-PET/CT of 93.8% is congruent with similar 68 Ga-PSMA studies (specificity > 90% for T3b disease) [12, 25, 26]. Therefore, we recommend nuclear medicine physicians to report on the presence of rT3a-b specifically.

Our study has inherent limitations. Since the PET/CT resolution is confined at 5 mm, limited diagnostic accuracy for small PCa-foci is to be expected. A selection bias has been introduced due to the selection of patients with biopsy-confirmed csPCa. It is thus unclear how 18F-DCFPyL-PET/CT performs in a truly biopsy naïve cohort of patients. The present study was set up to evaluate the capability of PSMA-PET/CT to guide targeted prostate biopsy for the detection of csPCa, not with the goal to discriminate between those who should be or should not be biopsied. Unfortunately, not all patients received a pre-operative mpMRI, limiting direct comparison to 18F-DCFPyL-PET/CT. Moreover, some of the patients who received a mpMRI had a longer interval between the mpMRI and the PET/CT scans due to the mpMRI being performed at the referring centre. Finally, since the RARP-specimen will always change shape (due to organ slicing and shrinking artefacts) when it is removed from the body, no truly exact anatomical correlation is possible. Therefore, the partial agreement score was used to correct for this pitfall.

Conclusions

When comparing the localisation of PCa on 18F-DCFPyL-PET/CT with the RARP specimen using anatomical mapping, an accurate per-patient localization (93%) of csPCa was found within the prostate. 18F-DCFPyL-PET/CT proves promising for PSMA-targeted biopsy and provides a moderate local staging ability.