Abstract
68Ga-labeled prostate-specific membrane antigen (68Ga-PSMA) PET/CT has a proven role in staging and restaging of prostate cancer (PCA). The aims of this study were to evaluate the association of intraprostatic 68Ga-PSMA PET/CT findings and PSMA expression in immunohistochemical staining and generate a cutoff value for differentiation between normal prostate (PN) and PCA. Methods: The data of 31 patients (mean age, 67.2 y) who underwent prostatectomy and preoperative PET were retrospectively analyzed. On PET, focally increased uptake in the prostate was suggestive of tumor. A region of interest was placed on the suggestive area to generate an SUVmax; a similar region of interest was placed on adjacent visually PN. Both PCA and PN were stained with monoclonal anti-PSMA antibody (clone 3E6, 1:100, M3620). Results: All intraprostatic PCA lesions on PET could be confirmed histopathologically. In PN sections (n = 31), median staining intensity was mild, median percentage of stained cells was 20% ± 14.24%, and median immunoreactive score (IRS) was 1. In PCA sections (n = 31), median IRS was 3, median staining intensity was strong, and median percentage of stained cells was 80% ± 16.46%. The mean SUVmax (±SD) of PCA (14.06 ± 15.56) was significantly higher than that of PN (2.43 ± 0.63; P < 0.001). Receiver-operating-characteristic curve analyses of the SUVmax of PCA, validated by immunohistochemical staining in 62 tissue samples, showed the best cutoff to be 3.15 (sensitivity, 97%; specificity, 90%; area under curve, 0.987). Applied to multifocal PCA, it resulted in sensitivity and specificity of 87% and 97% respectively. The mean SUVmax of PCA and PN for an IRS of less than 2 (n = 26; 2.52 ± 0.64) was significantly lower than the mean SUVmax for an IRS of 2 or more (n = 36; 12.38 ± 15.02; P < 0.001). The mean SUVmax was significantly lower in PCA samples with fewer than 50% stained cells (n = 30; 2.81 ± 2.35) than in samples with 50% or more (n = 32; 13.34 ± 15.55; P < 0.001). There was no correlation between the SUVmax of PCA and Gleason score (P = 0.54). Conclusion: This study showed that SUVmax on 68Ga-PSMA PET/CT correlates significantly with PSMA expression in primary PCA, enabling the detection of PCA with a high sensitivity and specificity.
Prostate cancer (PCA) is the most commonly diagnosed cancer and the second leading cause of cancer death among men in the western world. The lifetime probability of being diagnosed with PCA is 14% (1). The usual diagnostic tools for PCA include prostate-specific antigen testing, digital rectal palpation, transrectal ultrasound, prostate biopsy, and histopathologic examination (2–4). Additionally, further imaging techniques such as MRI, bone scintigraphy, CT, and PET/CT are used for staging primary PCA and restaging biochemical recurrences (2,5). Currently, multiparametric MRI is the imaging tool recommended for detection of primary PCA and subsequent biopsy. The MRI results are evaluated on the basis of the Prostate Imaging Reporting and Data System, which grades parameters such as T2-weighted imaging, diffusion-weighted imaging, dynamic contrast-enhanced imaging, and MR spectroscopy on a 5-point scale and describes the risk of PCA, its aggressiveness, its localization, and relevant incidental findings (6). Diagnostic reliability appeared to be highest for tumors in the peripheral and central zones but is limited for tumors in the transitional zone (5). Therefore, there is a need for a more reliable imaging modality that dependably discloses all parts of the prostate gland and can be used even in patients with contraindications to MRI.
Recent studies reported that 68Ga-labeled prostate-specific membrane antigen (68Ga-PSMA) PET/CT has excellent detection rates for lymph node metastases, skeletal metastases, local relapses, and soft-tissue metastases compared with other PET tracers such as 18F- and 11C-labeled choline derivatives (7–11). The sensitivity and specificity for detecting local PCA using 11C- and 18F-choline have been reported to be 73% and 91%, respectively (2).
PSMA is a transmembrane protein with significantly increased expression in the cells and metastases of PCA compared with normal prostate (PN) and other physiologically PSMA-expressing tissues such as the brain, the lacrimal and salivary glands, and the proximal tubules of the kidneys (9,10,12,13). PSMA expression correlates with higher serum levels of prostate-specific antigen and higher Gleason scores (GS) (8,14). The mean SUVmax of 68Ga-PSMA in PN is therefore usually 4 times lower than that in PCA (4).
However, to our knowledge, no study has specifically evaluated whether SUVmax on 68Ga-PSMA PET/CT correlates with PSMA expression rates and percentage of PSMA-positive cells on immunohistochemical staining. The aims of this monocentric study were to perform such an evaluation and to generate a cutoff SUVmax for differentiation between PN and PCA.
MATERIALS AND METHODS
Ethics, Data Search, and Patient Selection
This retrospective study was approved by the institutional ethics review board (EA4/039/17), and all subjects signed an informed consent form. To be included, the patients had to have received histopathologic, clinical, or biochemical confirmation of PCA; undergone elective radical prostatectomy within 3 mo after the PET/CT examination; and have tumor specimens available for reanalyses so that we could correlate the histopathologic findings with the imaging results. All patients who had been examined in the Department of Nuclear Medicine from May 2014 until October 2016 were selected from the 68Ga-PSMA PET/CT database. Forty-one of them had undergone PET/CT because there was a high degree of suspicion of, or histopathologically confirmed, PCA. Thirty-one of these 41 patients met the inclusion criteria (Fig. 1). At first, we looked at the PET/CT images to localize the intraprostatic PCA; later, we correlated this location with the corresponding tumor specimen slides. We also compared the preoperative MRI reports with PET/CT findings.
Study design.
Immunohistochemistry
Formalin-fixed, paraffin-embedded tissue sections 4 μm thick from the Institute of Pathology were reevaluated and used for subsequent immunostaining. In 31 patients, a sufficient archival tissue specimen was available for immunohistochemical staining. Routine hematoxylin- and eosin-stained sections were used for diagnosis and reevaluation. After being mounted on Superfrost Plus slides (Fisher Scientific), the paraffin sections were dewaxed and rehydrated to water by a series of graduated ethanol washes. For antigen staining, the sections were incubated for 20 min in a microwave oven (800 W) using ethylenediaminetetraacetic acid buffer (10 mmol/L; pH 8.0). Monoclonal anti-PSMA (clone 3E6, 1:100, M3620 [Dako]) was used, and the tumor sections were incubated with the antibody at room temperature for 1 h. Then, the sections were counterstained with hematoxylin and finally analyzed. The immunohistochemical results were reported as staining intensity and percentage of positively staining cells following the immunoreactive score (IRS) and modified with a 4-point IRS classification (Table 1) (15). The immunohistochemical analysis was performed by 2 independent investigators.
Four-Point IRS Classification
Imaging Protocol
68Ga was eluted from a 68Ge/68Ga generator (Eckert and Ziegler Radiopharma GmbH). PSMA-HBED-CC (ABX GmbH) was labeled with 68Ga. PET/CT imaging was performed 60.9 ± 26.13 min after intravenous injection of 117.23 ± 19.86 MBq (Table 2). A Gemini TF 16 PET/CT scanner (Philips) was used. The 3-dimensional acquisition mode was used for all PET scans. Axial, sagittal, and coronal slices were reconstructed (144 isotropic voxels 4 mm3 each) using the standard reconstruction algorithm. Before the PET scan, a low-dose CT scan was obtained for anatomic mapping and attenuation correction (30 mAs, 120 kVp). Each bed position was acquired for 1.5 min, with a 50% overlap.
68Ga-PSMA PET/CT Acquisition Characteristics and Findings
Image Analysis
The images were analyzed on an Extended Brilliance Workspace workstation (Philips). The scans were reread by 2 nuclear medicine clinicians with more than 10 y of experience in reporting on PET studies. Any focal prostatic uptake higher than uptake in the circumferential tissues was considered pathologic. In addition, SUVmax was measured in the nearest visually defined PN tissue adjacent to the primary tumor. For patients with a multifocal primary tumor, PN SUVmax was measured adjacent to the pathologic sample used for immunohistochemistry. Multifocal tumors detected on 68Ga-PSMA PET/CT were validated using pathology reports. Intraprostatic lesions were documented using the 39-sector scheme and later were correlated with the corresponding pathology findings (16).
Statistical Analysis
Data were analyzed using IBM SPSS Statistics 24 for Microsoft Windows. An explorative data analysis was used to calculate the mean SUVmax of PCA and PN and with respect to IRS (<2 and ≥2) and percentage of stained cells (<50% and ≥50%). The cutoff, sensitivity, and specificity of SUVmax were calculated by analyzing the receiver-operating-characteristic curve. After testing for normal distributions according to the Kolmogorov–Smirnov test, the Spearman ρ-test was used to analyze the correlation between SUVmax and IRS, percentage of stained cells, and GS. The Mann–Whitney U test was used to compare the mean SUVmax of PN versus PCA, of IRS < 2 versus IRS ≥ 2, and of <50% stained cells versus ≥50% stained cells. All statistical analyses were 2-sided, and P values of less than 0.05 were considered statistically significant.
RESULTS
Patients’ Data
From the 31 patients, 31 PCA samples and 31 PN samples were investigated. A PET-corresponding tumor was found for all 31 PCA samples (Fig. 1). The mean age of the patients at the time of the PET scan was 66.57 ± 8.77 y. The indication for PET was staging in 28 patients and restaging in 3 patients. Of the 3 restaged patients, one each underwent external-beam radiation therapy, brachytherapy, and androgen deprivation therapy. In all 3 patients, 68Ga-PSMA PET/CT was performed at least 3 mo after the treatments. All patients underwent radical prostatectomy after PET. None of the patients developed adverse events or clinically detectable pharmacologic side effects after injection of the 68Ga-PSMA.
The mean prostate-specific antigen level was 17.49 ± 20.81 ng/mL. The GS was 7 in 29% of the patients, 8 in 35.5%, and 9 in 25.8%. Detailed information about the patients’ characteristics is in Table 3. Fifteen patients had a unifocal tumor, and 16 patients had a multifocal tumor. Examples of unifocal and multifocal PCAs are shown in Figures 2 and 3, respectively. One patient (3.2%) had retroperitoneal lymph node metastases, 4 (12.9%) had pelvic lymph node metastases, and one (3.2%) had bone metastases (Table 2).
Patient Characteristics
68Ga-PSMA PET/CT images showing unifocal PCA with GS of 3 + 4 = 7. (A) Axial PET image. (B) Fused PET/CT image. SUVmax of tumor was 13.9, IRS was 2, and 80% of cells were stained.
68Ga-PSMA PET/CT images showing multifocal PCA in peripheral zone with GS of 5 + 5 = 10. (A and C) Axial PET images. (B and D) Fused PET/CT images. SUVmax of lesion in B was 84.3 and that of lesion in D was 5.7. IRS was 3, and 80% of cells were stained.
Immunohistochemistry
The mean tumor size documented in the pathology report was 29.4 ± 13.7 mm (range, 9–60 mm). In PN sections (n = 31), the median staining intensity was mild and the median percentage of stained cells was 20% ± 14.24%; the median IRS was 1 (range, 0–2). In PCA sections, the median IRS was 3 (range, 1–3), the median staining intensity was strong, and the median percentage of stained cells was 80% ± 16.46% (Fig. 4).
Examples of immunohistochemical staining of PNs and PCAs with IRSs of 2 and 3. Immunohistochemical staining was performed with monoclonal anti-PSMA (clone 3E6, 1:100, M3620). (A) PN (4 × 10, 10 × 10, 30 × 10). (B) IRS of 2 (4 × 10, 10 × 10, 30 × 10). (C) IRS of 3 (2 × 10, 4 × 10, 10 × 10).
SUV
SUVmax and uptake time, that is, the time between injection of 68Ga-PSMA and acquisition of PET images, did not correlate significantly (Spearman ρ; P = 0.963). The mean SUVmax (n = 31; 14.06 ± 15.56) was significantly higher in PCA than in PN (n = 31; 2.43 ± 0.63; P < 0.001) (Fig. 5).
Box plots of PNs in comparison to PCAs. (A) Mean SUVmax was 2.43 ± 0.63 in PNs (n = 31) and 14.06 ± 15.56 in PCAs (n = 31; P < 0.001). (B) Mean SUVmax was 2.52 ± 0.64 for IRS < 2 (n = 26) and 12.38 ± 15.02 for IRS ≥ 2 (n = 36; P < 0.001). (C) Mean SUVmax was 2.81 ± 2.35 for < 50% stained cells (n = 30) and 13.34 ± 15.55 for ≥ 50% stained cells (n = 32; P < 0.001).
Receiver-operating-characteristic curve analyses of the SUVmax of PCA, validated by immunohistochemical staining in 62 tissue samples, showed the best cutoff to be 3.15, resulting in a sensitivity of 97% and a specificity of 90% (area under the curve, 0.987) (Fig. 6). When this cutoff was applied to non–immunohistochemically validated foci in multifocal PCA, it resulted in a sensitivity of 87% and a specificity of 97% for 68Ga-PSMA PET/CT.
Receiver-operating-curve analysis. Cutoff of 3.15 for SUVmax yielded sensitivity of 97% and specificity of 90% (area under curve, 0.987).
SUVmax and Immunohistochemical Staining
There was no correlation between mean tumor size and SUVmax (Spearman ρ; P = 0.651). The mean SUVmax of PCA and PN was 2.52 ± 0.64 for IRS < 2 (n = 26) and 12.38 ± 15.02 for IRS ≥ 2 (n = 36) in (Fig. 5). There was a significant difference in SUVmax between IRS ≥ 2 and IRS < 2 (Mann–Whitney U test; P < 0.001). The mean SUVmax for fewer than 50% immunohistochemically stained cells (n = 30) was 2.81 ± 2.35, compared with 13.34 ± 15.55 for 50% or more (n = 32; P < 0.001) (Fig. 5). There was a significant difference in mean SUVmax between tumor specimens with more than 50% stained cells and fewer than 50% (Mann–Whitney U test; P < 0.001). The Spearman ρ test revealed a significant correlation between SUVmax and IRS (P < 0.001), as well as between SUVmax and percentage of stained cells (P < 0.001). The data are summarized in Table 4. The mean SUVmax was lower in patients with a GS of less than or equal to 8 (5.81 ± 4.7) than in patients with a GS of more than 8 (9.59 ± 14.9); however, there was no statistical correlation between SUVmax and GS (P = 0.54).
SUVmax vs. Immunohistochemistry
PET/CT and MRI
Of the 31 patients, 20 had preoperative MRI reports that could be retrieved from the hospital’s database. The median interval between PET/CT and MRI was 2 mo. Twenty primary PCAs were seen on both PET/CT and MRI. PET/CT and MRI showed concordant results for 12 (60%) of the 20 and discordant results in the other 8 (40%). Of the 8 patients in the discordant group, PET/CT showed multifocal PCA in 7 (87.5%) whereas MRI showed unifocal disease. In the 8th patient, MRI showed multifocal intraprostatic lesions whereas PET/CT showed unifocal disease. Furthermore, PET/CT showed 6 lymph node metastases and MRI showed none.
DISCUSSION
To the best of our knowledge, this was the first study to generate a cutoff SUVmax, validated by immunohistochemistry, for separating PCA from PN by 68Ga-PSMA PET/CT images acquired on a Gemini TF 16 scanner. This validated cutoff of 3.15 for SUVmax enables the diagnosis of PCA with a high sensitivity and specificity in both unifocal and multifocal disease. In a previous study by our group, an SUVmax of 3.2 based purely on imaging, without histopathologic confirmation, resulted in a sensitivity of 94.3% and a specificity of 100% for differentiation between PCA and PN (4). Retrospectively, these findings are completely in line with our new and immunohistochemically validated SUVmax cutoff.
Furthermore, Rahbar et al. documented a significant difference (P < 0.001) in median SUVmax between PCA (11.0 ± 7.8) and PN (2.7 ± 0.9). This result is similar to our result of 14.06 ± 15.56 and 2.43 ± 0.63 in PCA and PN, respectively (17). This high and specific tumor uptake occurs because PSMA, a folate hydrolase 1 or glutamate carboxypeptidase 2, is highly expressed in primary PCA and metastatic lesions (13,18,19). In our study, immunohistochemical PSMA staining was also more intense in PCA than in PN, confirming recent studies showing either weak or even absent immunoreactive staining in PN and hyperplastic prostate glands (7,12,20). Consequently, PSMA is a good target, and 68Ga-PSMA PET/CT thus yields images with a high target-to-nontarget ratio. This result is of special interest in the fast-emerging field of multimodal image–guided biopsy.
Transrectal ultrasound–guided biopsy, one of the standard clinical procedures, misses a significant number of PCAs in the ventral segment of the prostate gland and in extreme lateral positions in the peripheral zone and apex. Recent studies showed that transrectal ultrasound biopsy misses around 30%–45% of PCAs in these areas (21–23). In patients who have undergone multiple transrectal ultrasound–guided biopsies with negative results, MRI-guided biopsy achieved detection rates of 41%–59% (24–27). To circumvent the problem with transrectal ultrasound, multiparametric MRI has been proposed as an adjunct alternative. Primarily performed on patients with a precedent negative biopsy result, multiparametric MRI is used to localize the primary tumor, stage the disease, plan nerve-preserving radical prostatectomy, and monitor active surveillance. In experienced centers, multiparametric MRI–generated Prostate Imaging Reporting and Data System results reach sensitivities of 85%–90% and specificities of 62%–68% (28). Recent reviews on the performance of Prostate Imaging Reporting and Data System versions 1 and 2 have found a high discriminative ability for tumor detection (area under the curve, 0.96 in version 1 vs. 0.90 in version 2). In comparison, the area under the curve for PET in our study was 0.987 (29,30). This result could have been caused by some of the limitations of multiparametric MRI mentioned previously (28). Among them, one clinically relevant limitation is the low detection rate in small tumors and those with a GS of less than 7 (17).
In our current PET study, the SUVmax of PCA lesions was higher when the GS was over 8 than when it was lower or equal to; however, GS did not correlate with SUVmax (P = 0.54), as was recently also shown by Ceci et al. (9). In a recent paper by our group, GS tended to correlate with SUVmax in 60 tumor lesions (P = 0.071) (4). These findings are supported by histopathologic studies in which PSMA expression was usually shown to be higher in lesions with a higher GS (14). Future multicenter studies with larger patient populations will finally help define the exact correlation between GS and SUVmax in PCA.
Irrespective of a potential correlation between SUVmax and GS, the uniquely high target-to-nontarget ratio as represented by a high SUVmax and the specific binding of the PET tracer to PSMA, in combination with the 3-dimensional PET images, should bolster the concept of multimodal imaging and image-guided biopsy and stereotactic therapy. The retrospective data on MRI versus PET/CT of our study showed that PET information and MRI information are complementary in some patients. In this context, the combination of multiparametric MRI and 68Ga-PSMA PET, ideally performed within a single examination by 68Ga-PSMA PET/MRI, might become the future gold standard for localization and staging of PCA.
The retrospective and single-center design of our study bears a risk of biased results: a prospective multicenter study validating the clinical value of 68Ga-PSMA PET/CT or even 68Ga-PSMA PET/MRI in primary PCA should be performed in the future. Another limitation was the low number of patients. Finally, although the possibility of PSMA-negative PCA, potentially resulting in some false-negative 68Ga PSMA PET findings, is uncommon, it cannot be excluded (13,19).
CONCLUSION
This study showed that SUVmax on 68Ga-PSMA PET/CT correlates significantly with PSMA expression on primary PCA and can be used to detect and localize PCA with a high sensitivity and specificity.
DISCLOSURE
No potential conflict of interest relevant to this article was reported.
Footnotes
Published online Aug. 3, 2017.
- © 2018 by the Society of Nuclear Medicine and Molecular Imaging.
REFERENCES
- Received for publication April 24, 2017.
- Accepted for publication July 6, 2017.