Introduction

Despite substantial improvements in early detection and local treatment, prostate cancer still represents the third leading cause of cancer-related death in men after lung and colorectal cancer [1]. Depending on the individual risk category, the disease will recur in about 20% to 40% of all patients after initial curative treatment (radical prostatectomy or external beam radiation therapy) [2,3,4,5]. Following radical prostatectomy, biochemical recurrence is defined as two confirmed prostate specific antigen (PSA) values >0.2 ng/mL [6]. Salvage radiotherapy (sRT) is currently the treatment of choice for patients with early biochemical recurrence (EBR) after radical prostatectomy. Recent studies demonstrated improved efficacy when sRT was performed at PSA values <0.2 ng/mL [7,8,9]. However, in up to 30% of patients, sRT does not affect PSA levels, most likely due to extrapelvic localization of the recurrence [10]. Therefore, an imaging modality detecting the tumor localization in EBR even at very low PSA values would be desirable. So far, multiparametric magnetic resonance imaging (mpMRI) or choline positron emission tomography/computerized tomography (PET/CT) were most commonly used for recurrence detection in prostate cancer. Recently published data for combined whole body MRI with dedicated pelvic mpMRI showed an overall detection rate of 21% in patients with PSA levels between 0.05–56.12 ng/mL (median 0.36 ng/mL) [11]. Also the sensitivity of choline PET/CT is low in patients with PSA levels <1 ng/mL, with a mean detection rate of 20% [12, 13]. Therefore, a differentiation between local recurrence and oligometastatic disease in patients with EBR was almost impossible and inhibited a precise stratification to appropriate treatment modalities.

The 68Gallium labeled tracer targeting the prostate specific membrane antigen (PSMA) 68Ga-PSMA-11 used in combined PET/CT has significantly improved the tumor detection rate in patients with EBR, being as high as 55% in patients with a PSA of 0.2–0.5 ng/mL [14,15,16]. Its efficacy has also been shown in patients after primary external beam radiation therapy [17]. However, the proximity of the bladder to the prostate fossa still impairs the detection of local recurrence in the prostate bed. Increased soft tissue contrast in the pelvis using PET/MRI instead of PET/CT might further improve the detection rate for patients with EBR [18, 19]. Improved tumor localization in the initial staging has already been shown for 68Ga-PSMA-11 PET/MRI [20]. However, a drawback in pelvic PET/MRI is the fact that on commercially available systems bone information is not yet available for attenuation correction (AC). This is substantially affecting the accuracy of PET images within or close to bony structures, since the missing density in the MRI based AC map will lead to an underestimation of the PET activity [21]. Recently, it was shown, that the integration of time-of-flight (TOF) information can significantly reduce image artifacts due to metal implants or bony structures and improve image quality in simultaneous PET/MRI [22, 23]. Therefore, further improvements in the detection rate of patients with very low PSA values at scan might be expected using a PET/MRI with TOF capability for image analysis.

We retrospectively evaluated 68Ga-PSMA-11 TOF-PET/MRI in our single center cohort with regard to lesion detection rate in patients with biochemical recurrence after prostatectomy.

Patients and methods

Patients

This retrospective study included all patients who underwent 68Ga-PSMA-11 PET/MRI for biochemical recurrence after prostatectomy at the Department of Nuclear Medicine (UniversityHospital Zürich) between April and December 2016. The local ethics committee approved the study protocol and all patients gave a general written informed consent for retrospective use of their data (BASEC Nr. 2016–02230). Recorded patient characteristics included age at scan, preoperative prostate specific antigen (PSA) value and PSA value at scan. Additionally, initial tumor stage/ grade and Gleason score as well as the postoperative resection margin status (R0 vs. R1) and prior androgen-deprivation therapy (ADT) were assessed. We analyzed the detection of PSMA-positive lesions overall and according to the region of detection: prostate fossa, local and distant lymph nodes, bone lesions and visceral lesions. Lymph node regions were further divided into 1) pelvic, 2) para-aortic, 3) mediastinal/supraclavicular, and 4) axillary. Furthermore, the maximum standardized uptake value (SUVmax) and the applied amount of tracer were analyzed.

68Ga-PSMA-11 PET/MRI

All patients underwent a single injection of 68Ga-PSMA-11 (mean dose ± standard deviation, 123.34 ± 16.1 MBq, range 90–153 MBq). A clinical routine whole-body PET/MRI was performed 60 min after injection on a hybrid scanner (SIGNA PET/MR, GE Healthcare, Waukesha, WI, USA) used in previous studies at our department [24]. The scanner comprises a 3-T MR system with TOF-PET detector ring installed between the body and gradient coils. A whole-body MR localizer scan was started before acquisition of PET data. Subsequently a 3D dual-echo, spoiled gradient recalled echo sequence (LAVA-FLEX) for AC, and a PET emission scan was recorded in list mode. The default number of bed positions was six and the acquisition time per bed position was 2 min. The whole-body protocol furthermore included specific sequences covering the pelvis, including a high resolution T1-weighted LAVA-FLEX sequence, T2-weighted fast recovery fast spin-echo sequence in two planes and diffusion weighted images (b values: 0, 300, 1000). To reduce 68Ga-PSMA-11 activity in the bladder, ureters, and kidneys, furosemide was injected intravenously 30 min prior to the 68Ga-PSMA-11 injection (0.13 mg/kg), and patients were asked to void prior to the scan. For attenuation correction an atlas-based MR-AC was used for the head, and for the remaining body air, lung, and soft tissue were segmented using the DIXON LAVA-FLEX sequences generating a fat-water based attenuation correction map. Details regarding the MRI sequences are given in the supplements (Supplement Table 1). The protocol scan time was 30 min. A dual board-certified radiologist and nuclear medicine physician (R1), incorporating both the MRI and PET information, analyzed all images. To investigate the influence of the different imaging modalities additional readouts including only MRI sequences (MRIonly) and 68Ga-PSMA-11 PET with only water weighted axial DIXON LAVA-FLEX (PETonly) were performed. Furthermore, to analyze interreader variability 2 additional readouts by radiologist with 2 (R2) and 0.5 years (R3) of nuclear medicine experience were conducted. The readers extracted clinical information for every readout. For the PET/MRI readout only lesions with high suspicion for recurrence were considered positive: focal 68Ga-PSMA-11 uptake in the soft tissue of the prostate bed, lymph nodes with an SUVmax ≥ 3 and/or pathologically increased size (≥ 5 mm for perirectal nodes, ≥ 8 mm for iliac / retroperitoneal nodes, ≥ 1 cm for inguinal nodes), focal bone uptake with correlating bone marrow replacement or focal uptake with correlating soft tissue lesion.

Statistical analysis

Descriptive statistics were used to display patient data as median, interquartile range (IQR), or number (percent). The detection rate was plotted against the absolute prostate specific antigen (PSA) value in ng/mL at scan. Images were generated using GraphPad Prism version 7 (PraphPad Software, Inc. La Jolla, CA, USA). Agreement among observers was evaluated with R (v. 3.4.0. R Foundation for Statistical Computing, Vienna, Austria; package DescTools v. 0.99.20) using Fleiss’s kappa (κ) [25] for all readers and Cohen’s κ for each reader pair [26]. Ninety-five percent confidence intervals (CIs) are reported for κ values. Interpretation of κ and ICC was based on a classification provided by Landis and Koch: 0.0, poor; 0.0–0.20, slight; 0.21–0.40, fair; 0.41–0.60, moderate; 0.61–0.80, good; 0.81–1.00, almost-perfect reproducibility [27].

Results

A total of 56 patients were retrospectively analyzed. Patient characteristics are shown in Table 1. Median age at scan was 69 years (IQR: 11), and median preoperative PSA value was 9.3 ng/mL (IQR: 12.7). All patients underwent initial radical prostatectomy. Twenty-five patients (44.6%) presented with an initial tumor stage pT2. Twenty-six patients (46.4%) had a pT3 tumor and 1 patient (1.8%) had a pT4 tumor. The median Gleason score was 7 (IQR: 2). Twenty-three percent of all included patients were initially nodal positive (N1), primary metastatic disease (M1) was not reported in any patient. A postoperative positive resection margin was reported in 26 patients (46.4%). Non-imaging guided salvage radiotherapy of the prostate bed was performed in 27 patients (48.2%). Pre-scanning ADT was administered in 8 patients (14.2%). Seven of these patients had intermittent ADT several years before the scan. Only 1 patient had continuous ADT at the time of the 68Ga-PSMA-11 PET/MRI. The total scan time per patient ranged from 25 to 57 min, with a median scan time of 42 min.

Table 1 Patient characteristics
Full size table

The 68Ga-PSMA-11 PET/MRI, the MRIonly and the PETonly findings of all 56 patients listed by increasing PSA values at scan are displayed in Table 2. In two patients, MRIonly was not evaluated, since the protocol was not complete. Overall, in 44 of 56 patients (78.6%) PSMA-positive lesions were detected using 68Ga-PSMA-11 PET/MRI. Detection rates subdivided by the PSA value at scan and the location of recurrence are shown in Fig. 1.

Table 2 Detection pattern by 68Ga-PSMA-11 PET/MRI (MRIonly / PETonly)
Full size table
Fig. 1
figure 1

Detection rate by 68Ga-PSMA-11 PET/MRI. a 68Ga-PSMA-11 PET/MRI detection rate stratified by different PSA levels at time of scan. b Number of 68Ga-PSMA-11 PET/MRI positive events stratified by side and PSA levels at time of scan

Full size image

The detection rate of MRIonly was significantly lower compared to PET/MRI with 13 of 54 patients (24%). The overall detection rate of PETonly was comparable with PET/MRI with 43 of 56 patients (76%). However, six lesions suspicious on PETonly were interpreted as physiologic activity on PET/MRI and 6 suspicious lesions on PET/MRI were missed on PETonly. MRIonly was significantly inferior to PET/MRI and PETonly especially for the detection of lymph node and bone metastases (Supplement Table 2).

The detection rate was 44% (four of nine patients) for PSA values of 0 to < 0.2 ng/mL. Suspicious lesions were detected in pelvic (n = 2) and para-aortic (n = 1) lymph nodes and in the bone (scapula, sternum, n = 2). In patients with a PSA between 0.2 and < 0.5 ng/mL, the detection efficacy increased up to 72.7% (eight of 11 patients). Positive lesions were detected in the prostate fossa (local recurrence; n = 2), in pelvic (n = 5), and para-aortic (n = 1) lymph nodes and in the sacrum (n = 1). In 12 of 15 (80%) patients with a PSA between 0.5 and < 2.0 ng/mL, suspicious lesions were detected. Sites of detection were the prostate fossa (local recurrences; n = 4), pelvic (n = 8), and para-aortic (n = 1) lymph nodes, as well as bone metastases in four patients. In 20 of 21 patients (95.2%) with a PSA >2.0 ng/mL at scan, at least one lesion was PSMA-positive. Three of these patients presented only local recurrence at scan time (PSA 4.1, 4.5, and 7.6 ng/mL); one patient demonstrated additional PSMA-positive pelvic lymph nodes (PSA 2.2 ng/mL). Lymph node metastases were found in 16 of these patients: seven patients presented pelvic lymph nodes only, three patients presented pelvic and para-aortic lymph nodes, four patients presented pelvic, para-aortic and mediastinal/supraclavicular lymph nodes, one patient presented pelvic and axillary lymph nodes. In four patients, additional bone metastases were found. Bone marrow abnormalities were detected in all sites of suspicious PSMA uptake (Supplement Figure 1).

One patient presented with a single PSMA-positive lesion within the liver (PSA 30 ng/mL). Subsequently performed liver MRI confirmed a lesion of the falciform ligament that was further investigated with laparoscopy and histologically confirmed as peritoneal carcinomatosis. Interestingly, one patient with a PSA of 4.65 ng/mL at scan time was completely PSMA-negative. Re-imaging was performed using 18F–Choline PET/CT to overcome a possible PSMA-negative tumor. However, also 18F–Choline PET/CT could not reveal any metastatic lesion. Exemplary images of local recurrence, lymph node and bone metastases are shown in Figs. 2, 3, 4, and 5.

Fig. 2
figure 2

Local recurrence in 68Ga-PSMA-11 PET/MRI. a MIP 68Ga-PSMA-11 PET of a eighty-four year old patient with a single PET/MRI positive lesion of 1.4 × 2.0 × 2.3 cm, SUVmax 19.7 between bladder neck and rectum (PSA 4.1 ng/ml). Radical prostatectomy had been performed 18 years ago. Initial tumor stage was pT2b, pN0 (0/9), cM0, G2, Gleason 3 + 3 = 6, R1. No signs of lymph node or bone metastases were detectable. b axial T2 weighted image shows obvious soft tissue lesion (arrow), with good correlation to the increased 68Ga-PSMA-11 uptake on fused PET/MR images c). d The lesion can be seen also on the axial water weighted DIXON LAVA-FLEX sequence

Full size image
Fig. 3
figure 3

Small PSMA-positive local recurrences in low PSA values (~0.5 ng/mL). a MIP 68Ga-PSMA-11 PET of a 71-year-old patient with a PSA level of 0.49 ng/mL at scan time. b Unclear uptake on axial 68Ga-PSMA-11 PET. c No suspicious focus on T2 axial imaging, d however clear projection of small focal PET-positive lesion (SUVmax: 4.4) in the right dorso-lateral prostate fossa (arrow), suspicious for local recurrence. e MIP 68Ga-PSMA-11 PET of a 74-year-old patient (PSA 0.51 ng/mL at scan) with initial pT2c, pN0, M0, Gleason score 3 + 3 = 6. Similar findings of f unlear 68Ga-PSMA-11 uptake, g no suspicious finding on anatomical imaging, however h clear projection of PET-positive lesion (SUVmax: 13) into the soft tissue, indicating local tumor recurrence

Full size image
Fig. 4
figure 4

PSMA-positive lymph node metastases. a Axial 68Ga-PSMA-11 PET image in a patient 3 years after radical resection (pT2c, pN0 (0/12), cM0, Pn1, Gleason 4 + 4 = 8), with a current PSA value of 0.22 ng/mL. Focal uptake (SUVmax: 6.4) in the left perirectal region, b projecting onto a small lesion on fused axial PET/MRI, c corresponding to a small (3 mm) lymph node (arrow) on axial T1 weighted image. d Axial 68Ga-PSMA-11 PET image in a patient 8 years after resection (pT2c, pN0 (0/23), cM0, Gleason score 3 + 4 = 7, R0) with focal uptake in the left intern iliac region (SUVmax: 9.7). e Fused PET/MRI images and f corresponding 8 mm lymph node (arrow) on axial T1 weighted image

Full size image
Fig. 5
figure 5

PSMA-positive bone metastases in patients with low PSA values. a MIP 68Ga-PSMA-11 PET of a 65-year-old patient with a solitary 1.1 cm PSMA-positive (SUVmax: 4.1) bone metastasis in the sternum. PSA value at scan 0.14 ng/mL. b Coronal PET image with small focal uptake and c coronal T2 weighted MRI shows focal hypointensity in the sternum (arrow), d correlating well with the increased PSMA uptake. Five months after stereotactic radiotherapy the PSA value dropped to 0.04 ng/ml. Initial tumor stage was pT2c, cN0, cM0, Gleason 3 + 4 = 7, R1. e/f MIP and axial 68Ga-PSMA-11 PET with a 68Ga-PSMA-11-positive osseus lesion in the right acetabulum (SUVmax 5.9) g focal hypointensity on the T1 weighted axial image (arrow), h correlating well with the increased PSMA uptake on fused PET/MR images. PSA value at scan 0.96 ng/mL

Full size image

Median SUVmax was 11.4 (IQR: 8.5) in local recurrence and 9.4 (IQR: 16) in bone metastases. In PSMA-positive lymph nodes a median SUVmax of 9.7 (IQR: 12.3) corresponded to a median node size of 6 mm (IQR: 4).

The overall interreader agreement for the three readers for cancer detection was excellent (kappa = 0.796, CI 0.645–0.947), with almost perfect agreement for bone metastasis (kappa 0.924), and substantial agreement for lymph node metastasis (kappa = 0.697, CI 0.556–0.838) and local recurrence (kappa = 0.66, CI 0.522–0.808). The pairwise interreader agreement is given in the Supplement Table 3, showing an almost perfect agreement between R1 and R2.

Discussion

In our current patient population 68Ga-PSMA-11 PET/MRI revealed a high overall detection rate of 78.6% in patients with biochemical recurrence after radical prostatectomy. The detection rate was only slightly lower for PSA values between 0.2 and < 0.5 ng/mL at scan. These results underline the promising potential of 68Ga-PSMA-11 PET/MRI as re-staging tool for prostate cancer patients with early biochemical recurrence and low PSA values.

A direct comparison of our overall detection rate to other reports in the literature is difficult owing to the low number of series combining 68Ga-PSMA-11 PET with MRI and owing to different patient characteristics at the time of scan [18, 19, 28]. Afshar-Oromieh et al. reported a comparable overall detection rate of 80% in their initial experience including 20 consecutive patients [18]. However, the median PSA value at scan in their cohort was considerably higher, with 2.62 ng/mL (range of 0.51–73.60 ng/mL) compared to a median PSA of 0.99 ng/mL (range of 0.05–30 ng/mL) in our cohort. A study including 119 patients undergoing simultaneous 68Ga-PSMA-11 PET/MRI and PET/CT showed an overall PSMA-positive rate of 78.2% [19], again with a higher median PSA value of 1.7 ng/mL (confidence interval: 1.25–2.2) at scan. The same group reported a similar accuracy in detecting lymph node and bone metastases using a non-TOF 68Ga-PSMA-11 PET/MRI compared to PET/CT [28]. The largest series of patients published so far by Afsahr-Oromieh et al. including 1007 patients had an overall detection rate of 79.5% with a median PSA of 2.2 (range 0.01–1′237 ng/mL) [29].

The detection rate was significantly lower for the MRIonly readout in our cohort (24%), which is in line with first results from Robertson et al. (21%) [11]. However, the present protocol is not a clinically acceptable whole body MRI, with high resolution images in only one plane, and no whole body DWI. Therefore, we would like to emphasize that our results of the MRIonly readout probably underestimate the potential of whole body MRI and were only performed to gauge the information-gain from the added PSMA-PET. Only one lesion was rated suspicious on MRIonly and negative on PET/MRI or PETonly. After interpretation of the MRI images with the PET the lesion was interpreted rather as scar tissue by R2. PETonly had a superior detection rate compared to MRIonly with only six false negative and eight false positive lesions.

The interreader agreement between R1 and R2 was almost perfect and good between R1 and R3. R3 has only a limited experience with nuclear medicine readouts. These results are in line with the interobserver assessments by Fendler et al. for 68Ga-PSMA PET/CT and suggest that the interpretation of PET/MRI is as robust as the interpretation of PET/CT [30].

The most challenging and clinically important cohort are patients with PSA values <0.5 ng/mL at the time of scan. These patients might benefit from a targeted salvage radiotherapy or local surgery of oligo-metastatic disease at the site of recurrence [7, 31]. In our cohort, 25% of the PSMA-positive recurrences in patients with PSA values <0.5 ng/mL were located outside the pelvis. Local recurrence alone was seen only in 10% of patients in this subgroup, while 35% of patients presented with pelvic lymph node metastases and 15% with bone metastases. Therefore, most of these patients probably would not benefit from a non-targeted salvage radiotherapy to the prostate fossa alone, which is, in the absence of imaging-detected metastases, still standard of care in patients with EBR following radical prostatectomy [3].

The detection rate in our cohort was 72.7% in patients with low PSA values between 0.2 and <0.5 ng/mL. The hitherto highest reported detection rates in patients with the same PSA range using 68Ga-PSMA-11 PET/CT by two different groups were 57.9% and 50% [14, 15, 29], respectively. Furthermore, a detection rate of 44% has been reported in 27 patients with a PSA <1 ng/mL, and of 50% in 22 patients with PSA values between 0.2 and 2 ng/mL [32, 33]. The largest cohort published had a detection rate of 46% in 108 patients with PSA values between 0.2 and < 0.5 ng/mL [29]. However, the patient number in most of the aforementioned studies and in our own cohort are rather low, making it difficult to draw a definitive conclusion.

Our data show a local recurrence in 17.8% (10/56) of the patients. However, only two of these patients had a PSA <0.5 ng/mL at scan. Five patients with a PSA <0.5 ng/mL presented a PSMA-positive extra pelvic focus. Other groups reported a local recurrence in 35.1% (n = 248), 15.1% (n = 119), 21.6% (n = 37), and 0% (n = 20) of their patients [15, 18, 19, 34].

PSMA-positive lymph nodes were detected in 57.1% of our patients, whereas bone lesions were detected in 19.6%.These data are comparable to the collective of Eiber et al., who reported PSMA-positive lymph nodes in 57.6% and bone lesions in 35.9% [15]. An association between rising PSA values and PSMA-positive lymph nodes is not surprising. Yet, the three PSMA-positive bone lesions in patients with a PSA <0.5 ng/mL have to be highlighted, since bone lesions are usually not expected in patients with such low PSA values.

68Ga-PSMA-11 is almost completely excreted renally and several studies demonstrated a significant halo-effect impairing the image quality around the kidneys and the bladder [35,36,37]. Most centers use a 60 min interval between tracer injection and image acquisition [37,38,39,40]. Attempts to reduce artifacts due to high tracer accumulation in the bladder and the kidneys include wash-out of the tracer by using diuretics, early imaging after minutes or very late imaging after 3 h [16, 40]. The optimal time point for imaging is controversial. Early imaging will give the opportunity to image the prostate bed without spillover from urine activity, while late scans improve the tumor to background ratio. Clear anatomical allocation of tracer activity within the prostate bed, which is feasible with simultaneous PET/MRI, might reduce the interference of urine activity with local recurrence (Fig. 3).

Another, recently reported possibility to reduce artifacts on PET/MRI is the use of TOF information for image reconstruction [22]. The narrowed decay allocation on the line of response due to TOF capacity in PET/MR scanners can improve image resolution and reduce artifacts resulting from incorrect AC [41]. This can especially reduce arm truncation artifacts leading to substantial image artifacts on non TOF PET/MRI [42]. Segmentation for MR-AC is still performed without the incorporation of bone density in most PET/MRI systems. This leads to a systematic underestimation of PET activity around osseous structures [21].

Clinical feasibility for PET/MRI scans is another important fact. With a protocol length of 30 min a median scan time of 42 min with a maximum of 57 min was achieved. Reflecting the fact that some patients needed a toilet break, or there were difficulties with planning the scan.

Other factors such as continued androgen deprivation therapy might also influence the detection rate of 68Ga-PSMA-11 imaging. In vitro data demonstrated an upregulation of PSMA surface receptors by androgen deprivation therapy [43]. In our cohort, only one patient underwent 68Ga-PSMA-11 PET/MRI during androgen-deprivation therapy with 68Ga-PSMA-11 positive lymph nodes at a PSA level of 0.08 ng/mL. Also Afsahr-Oromieh et al. found a significant correlation between ADT at PET scan and pathological findings on PSMA PET/CT [29]. Other, hitherto published studies do not mention if androgen-deprivation was given before imaging.

A limitation of our study is its retrospective nature, leading to an inherent selection bias. Furthermore, the reported patient collective is relatively small and lacks a comparison to another imaging modality (e.g. PET/CT) and histopathology of PSMA-positive lesions. Future investigations are needed to elucidate whether PET/MRI can improve the detection rate of 68Ga-PSMA in patients with early biochemical recurrence and furthermore whether targeted treatment of oligo-metastatic disease improves the cancer-specific survival.

Conclusion

Our data shows that simultaneous 68Ga-PSMA-11 PET/MRI with TOF has a high detection rate for biochemical recurrence. Furthermore, it shows that even at very low PSA levels <0.5 ng/mL, extrapelvic disease can be localized in 25% of patients with EBR, potentially changing the sRT-planning in a substantial number of patients.