Comparison of 68Ga-PSMA-617 PET/CT with mpMRI for the detection of PCa in patients with a PSA level of 4–20 ng/ml before the initial biopsy

The study was aimed at assessing the diagnostic performance of 68Ga-PSMA-617 PET/CT in the detection of prostate cancer (PCa) in patients with a prostate-specific antigen (PSA) level of 4–20 ng/ml and to compare its efficacy with that of multiparametric MRI (mpMRI). We analyzed the data of 67 consecutive patients with PSA levels of 4–20 ng/ml who almost simultaneously underwent 68Ga-PSMA-617 PET/CT and mpMRI. 68Ga-PSMA-617 PET/CT and mpMRI diagnostic performances were compared via receiver operating characteristic (ROC) curve analysis. Of the 67 suspected PCa cases, 33 had pathologically confirmed PCa. 68Ga-PSMA-617 PET/CT showed a patient-based sensitivity, specificity, and positive and negative predictive values (PPVs and NPVs) of 87.88%, 88.24%, 87.88%, and 88.24%, respectively. The corresponding values for mpMRI were 84.85%, 52.94%, 63.64%, and 78.26%. The area under the curve values for 68Ga-PSMA-617 PET/CT and mpMRI were 0.881 and 0.689, respectively. 68Ga-PSMA-617 PET/CT showed a better diagnostic performance than mpMRI in the detection of PCa in patients with PSA levels of 4–20 ng/ml.

Imaging analysis. The 68 Ga-PSMA-617 PET/CT images were reviewed by consensus between three experienced nuclear medicine physicians by using Siemens MMWP workstation. PCa lesions were distinguished from the surrounding normal prostate tissues by visual analysis. To calculate the 68 Ga-PSMA-617 uptake of the primary PCa lesion, circular regions of interest (ROIs) were drawn around areas with focally increased uptake in transaxial slices by using e.soft software (Siemens) at a 30% threshold, and SUVmax was calculated automatically. Scans were considered positive when the focal uptake of 68 Ga-PSMA-617 was superior to the background activity. Semiquantitative measures comprised SUVmax and SUVratio (explained below under "Statistical analysis"). The positive criterion for mpMRI is an abnormal focal signal increase or decrease. The readers were instructed to decide on the basis of their overall impression by using the 5-point Prostate Imaging Reporting and Data System (PI-RADS) score 19 : 1, PCa highly unlikely; 2, PCa unlikely; 3, equivocal PCa; 4, PCa likely; and 5, PCa highly likely. At the patient level, data were dichotomized (score of 1-2, PCa negative; score of 3-5, PCa positive) using the highest score per patient.  16,20 , repeated saturation biopsy (≥ 18 cores) was performed in negative cases during the follow-up to avoid a potential false-negative diagnosis in the initial biopsy. A diagnosis of malignancy was defined as a positive biopsy result. Benign diagnoses were comprehensively defined as repeated negative pathological findings on biopsies and continuous follow-up by monthly PSA monitoring, imaging-based monitoring (mpMRI or 68 Ga-PSMA-617), and clinical symptom examination for at least 1 year. Subsequently, PCa patients who met the indications for surgery received robot-assisted laparoscopic radical prostatectomy (RALRP) and prostatectomy specimens were obtained. The biopsy tissues and specimens were formalin-fixed and routinely processed by hematoxylin-eosin staining and immunohistochemistry (IHC) analysis as previously reported 21 . Histopathological results of the biopsy tissues served as a reference for PCa diagnosis, which was stratified according to the 8th edition of the AJCC staging system for PCa 22 .
Statistical analysis. Continuous variables are presented as means ± standard deviations or medians (interquartile ranges). Categorical variables are presented as frequencies (percentages). The number of PCa cases determined using 68 Ga-PSMA-617 and mpMRI were counted, and the diagnostic performance of 68 Ga-PSMA-617 PET/CT and mpMRI was evaluated. The overall diagnostic accuracy of the two methods by using patient-level data was assessed by receiver operating characteristic curve (ROC) analysis. Areas under the ROC curves (AUCs) with 95% confidence intervals (CIs) were compared to each other as well as to a fixed value of 0.5. SUVratio was determined by dividing the highest SUVmax in PCa lesions by the SUVmax of the background (maximal uptake in the normal tissues surrounding the prostate gland), which reflected the imaging contrast of 68 Ga-PSMA-617 PET/CT. We used one-way ANOVA and Student's t test to evaluate the relationships between various SUVmax or SUVratio measurements and clinical parameters such as age, PSA value, and GS. P values < 0.05 were considered statistically significant. Correlation analysis was assessed using the Spearman rank correlation coefficient. The provided boxplots show the first and third quartiles and median. The ends of the whiskers represent the 10th and 90th percentiles. The data were analyzed using IBM SPSS software, version 24.0 (IBM, Inc., Chicago, IL, USA), GraphPad Prism software, version 7.0 (GraphPad Software, Inc., La Jolla, CA, USA), and R software, version 3.4.2 (R Foundation for Statistical Computing, Vienna, Austria).

Results
Patient characteristics. In this study, 115 men underwent 68 Ga-PSMA-617 PET/CT and mpMRI, among whom 67 patients were included in the final analysis, according to the inclusion and exclusion criteria shown in Fig. 1. The mean time between the two examinations ( 68 Ga-PSMA-617 PET/CT and mpMRI) and the histopathological examination was 17.5 days (range 1-56 days) and 15.64 days (1-35 days), respectively.
Semiquantitative analyses and optimal cut-off points of ROC curves: detection of PCa and clinically significant PCa. On ROC analysis (Fig. 4A), we determined the optimal cut-off points for SUVmax, SUVratio, and PI-RADS score according to the results of pathology examinations. We investigated the data by categorizing the lesions as benign (benign prostatic tissue) or malignant (any GS  19 , the cut-off PI-RADS score was selected as 3, with AUC = 0.7210 (95% CI 0.6190-0.8090), which was significantly different from the 68 Ga-PSMA-617 uptake values (SUVmax or SUVratio) (P1 = 0.0427, P2 = 0.0013).
We further investigated the data by dividing the lesions into non-clinically significant PCa (PCa with a Gleason score of 6 and benign prostatic tissue) versus clinically significant PCa (PCa with GS ≥ 7). The ROC analysis demonstrated that the optimal cut-off points were 5.050 (SUVmax) and 2.365 (SUVratio), which could differentiate tumor lesions from benign conditions with sensitivities of 83.33% and 86.67% respectively, and specificities of 91.89% and 97.30%, respectively, thereby indicating the largest AUC (AUC1 = 0.9086, 95% CI 0.8308-0.9864;  www.nature.com/scientificreports/ AUC2 = 0.9477, 95% CI 0.8922-1) (Fig. 4B). The ROC analysis by statistical software demonstrated that when the PI-RADS score was 3.5, the AUC was the largest (AUC = 0.8095, 95% CI 0.7086-0.9103). When the PI-RADS score was 3 or 4, the corresponding AUC was 0.689 or 0.761, respectively, and there was no statistical difference between the two results (P = 0.2492), while the SUVratio showed a significant difference (P = 0.0093) and the SUVmax showed no statistical difference (P = 0.1164).
comparison of the clinical utility between 68 Ga-PSMA-617 PET/CT and mpMRI. As shown in Fig. 5, to determine which diagnostic parameters of 68 Ga-PSMA-617 PET/CT and mpMRI had better clinical utility in assisting biopsy decisions, we performed decision curve analysis (DCA) of SUVmax, SUVratio, and PI-RADS score. The grey line (leftmost) represents the "biopsy all patients" strategy, and the horizontal black line indicates the "biopsy none" strategy. Curves representing each diagnostic parameter are indicated. As expected, all of the methods were superior to the "biopsy all patients" strategy, and the net benefit of PI-RADS was the lowest of the three methods, while the net benefit of SUVratio was higher than that of SUVmax. The net benefit of SUVmax and SUVratio was greater for patients within threshold probabilities of 20-95% than PI-RADS, with the outcome as PCa (Fig. 5), suggesting that the clinical utility of 68 Ga-PSMA-617 PET/CT was superior to that   Table 4 and SI Figure S3, we have already confirmed the positive correlation between SUVmax and GS. SI Figure S5 intuitively showed this correlation. Furthermore, to evaluate the expression pattern of PSMA in PCa, we performed immunohistochemical staining in PCa tissues from 25 patients who agreed to provide tissue specimens. Among the 25 samples, one (4%) did not show detectable PSMA expression, while three (12%) samples showed PSMA expression in < 50% of PCa tissues, and in 21 (84%) samples, PSMA expression was > 50%.

Discussion
To the best of our knowledge, this is the first study to evaluate the diagnostic performance of 68 Ga-PSMA-617 PET/CT and mpMRI in the detection of PCa with PSA levels of 4-20 ng/ml. Our data support the hypothesis that 68 Ga-PSMA-617 PET/CT can more accurately differentiate primary PCa with PSA levels of 4-20 ng/ml than mpMRI, and that 68 Ga-PSMA-617 uptake is positively correlated with the clinical risk parameters GS and PSA level. Moreover, as an important parameter of 68 Ga-PSMA-617 PET/CT, the SUVratio seems to be a superior predictive factor owing to its better discriminative ability to predict PCa, thereby improving diagnostic performance. Considering the clinical difficulties in the diagnosis of PCa with PSA levels of 4-20 ng/ml, serum PSA screening might not seem to be an ideal method for the detection of PCa. Lojanapiwat et al. reported that when the cut-off PSA level was higher than 20 ng/ml, the specificity was significantly improved 23 . In comparison with traditional imaging studies, mpMRI is currently the most commonly used clinical imaging examination method, through T1, T2-weighted images, DWI, DCE, and other functional sequences. Multi-analysis imaging can better show the integrity of the prostate capsule and tumor invasion into tissues and organs around the prostate. MpMRI can also show invasion of pelvic lymph nodes and lesions of bone metastases, which may play important roles in clinical staging. Current guidelines propose the use of mpMRI for local staging in primary PCa 24 . However, clinical MRI examinations cannot simultaneously perform a whole-body scan and usually perform well for local conditions such as those limited to the chest, abdomen, pelvis, etc. Whole-body assessments using these examination methods have limitations, making the examination less comprehensive and the staging not accurate enough. In addition, it is difficult for elderly patients with PCa to remain motionless for a long time during the examination process, thus resulting in motion artifacts. Furthermore, metal implants such as pacemakers and artificial joints are contraindicated in this examination. In contrast, 68 Ga-PSMA-617 PET/CT is a non-invasive and full-body imaging modality, with fewer contraindications and easier acceptance by patients 25 . www.nature.com/scientificreports/ Our study systematically evaluated the potential of 68 Ga-PSMA-617 PET/CT for the detection of primary PCa with PSA levels of 4-20 ng/ml, and the findings indicated the excellent diagnostic performance of 68 Ga-PSMA-617 PET/CT in comparison with mpMRI. In terms of sensitivity, 68 Ga-PSMA-617 PET/CT and MRI both showed good diagnostic performance (87.88% vs. 84.85%), while in terms of specificity, 68 Ga-PSMA-617 PET/ CT was better than mpMRI (88.24% vs. 52.94%). In their study based on histopathological segment analysis, Eiber et al. 26 demonstrated that in comparison with mpMRI, 68 Ga-PSMA PET/CT had higher diagnostic accuracy (AUC), sensitivity, and specificity for the detection of primary PCa (0.83 vs. 0.73, 92% vs. 66%, and 94% vs. 82%, respectively). Giesel et al. 27 , Scheltema et al. 28 , and Chen et al. 29 also reported that 68 Ga-PSMA PET/CT offers advantages over mpMRI for intra-prostatic tumour localization. These previous results were consistent with our present study in terms of the diagnostic performance of 68 Ga-PSMA PET/CT; however, for mpMRI, our study showed a lower specificity. A possible reason for this discrepancy might be that the definition of PCa in our present study included GS ≥ 6 and the diagnostic selected threshold of PI-RADS was 3. When we only considered clinically significant PCa (GS ≥ 7) and chose PI-RADS 4 as the diagnostic selected threshold, the diagnostic accuracy (AUC) of mpMRI went up to 0.761, which was consistent with the previous studies, but still significantly lower than that of 68 Ga-PSMA PET/CT. Our result also contradicted the findings reported by Zhang et al. 30 and Zang et al., 31 who showed that the detection rate of 68 Ga-PSMA PET/CT in PCa was similar to that of mpMRI in Chinese patients. This was mainly because our study assessed PCa patients with PSA levels of 4-20 ng/ml and mainly focused on intra-prostatic tumours rather than metastatic PCa lesions or patients with PSA levels ≥ 20 ng/ml. Therefore, our study is different from previous studies in terms of the methods employed and/or the patient range, and the diagnostic accuracy of 68 Ga-PSMA PET/CT and mpMRI observed in the present study are more applicable to patient groups with low-intermediate risk.
In contrast to the organ specificity of PSA, PSMA is a transmembrane glycoprotein expressed on the cell surface of PCa cells at a much higher concentration than in normal prostate cells 11,32 , and though there were a few benign lesions of hyperplasias or inflammations, with a small number of false-positive cases, the overall image results were more theoretically reasonable than those obtained with mpMRI for detection of PCa with PSA levels of 4-20 ng/ml. Without consideration of the limiting PSA cut-off (20 ng/ml), a previous systematic review and meta-analysis by Perera et al. 14 reported that the summary sensitivity and specificity of 68 Ga-PSMA-11 PET/CT were both 86% in the patient-basis analysis of PCa, and the present study showed slightly higher results for sensitivity and specificity (87.88% and 88.24%). Thompson et al. 9 and Kim et al. 10 showed that the specificity of mpMRI was only 36-58% in PCa diagnosis, indicating a high false-positive rate, which was similar to the findings of the present study (52.94%). Although another systematic review and meta-analysis showed that the pooled average sensitivity and specificity of mpMRI for PCa detection were 89% and 73% 8 , it precisely confirmed the controversial diagnostic performance of mpMRI. These results indicated that for patients with PSA levels of 4-20 ng/ml, 68 Ga-PSMA-617 PET/CT might possess greater ability and stability to identify benign prostatic tissue than mpMRI to avoid more false-positive cases, which might reduce overdiagnosis.
The density of PSMA receptors was reported to increase with more aggressive PCa cells 33,34 ; therefore, 68 Ga-PSMA-617 PET/CT could be used to identify higher-grade malignancies of PCa in a cohort with GS. The diagnostic accuracy of PET is known to be dependent on the harmonization of 68 Ga-PSMA-617 PET/CT scanning and the expertise of the reporter, and on developing quantitative aids that allow the reader to determine likely clinical significance 28 . We found that the semi-quantitative SUVmax and SUVratio of lesions would be appropriate parameters to detect PCa lesions in 68 Ga-PSMA-617 PET/CT, and were superior to the PI-RADS score in mpMRI. In our study, we found a significant correlation between higher GS of PCa lesions and higher SUVmax and SUVratio values, which is consistent with the findings of the previous study 35 . And we defined 2.040 and 4.795 as the best cut-off values for SUVmax (sensitivity 81.82%, specificity 91.18%, AUC = 0.9180) and SUVratio (sensitivity 90.01%, specificity 94.12%, AUC = 0.9608), respectively, for PCa detection. Lopci et al. 25 reported that, for patients with overall PSA level, the ROC analysis demonstrated that a SUVmax > 4.8 could differentiate tumour lesions from benign conditions with a sensitivity of 82.4% and specificity of 72.2% (AUC = 0.843). The corresponding cut-off SUVratio was > 1.8, for which sensitivity was 94.1% and specificity was 88.9% (AUC = 0.949). The discrepancy between our own cut-off values and the values reported previously may be primarily attributable to the small sample size of the studies and to technical differences in scanners and reconstruction modalities. In comparison with a previous study, our results showed a slightly higher specificity. The reason was that the calculation in the previous study was performed in benign prostatic hyperplasia (BPH) 25 , which was characterized by slightly higher uptake values than the normal prostate corresponding to our study. Like the previous study reported 15 , our results showed that the clinical utility and benefit of 68 Ga-PSMA-617 PET/ CT was superior to that of mpMRI by decision-curve analysis, and SUVratio might be the preferred diagnostic parameter in the detection of PCa with PSA levels of 4-20 ng/ml.
In our study, the sensitivity of 68 Ga-PSMA-617 PET/CT was relatively higher than that of 68 Ga-PSMA-11, in comparison with previous studies 14,35,36 . Using a different radio-tracer, different patient selection criteria or randomly higher PSMA expression of the enrolled PCa patients are possible reasons. Meanwhile, the high PSMA expression of the enrolled PCa patients might be another possible reason. Immunohistochemical analysis was performed on 25 PCa tissue samples, and the PSMA-positive rate was up to 96%. Furthermore, the pathological findings of the present study indicated that all tumors were possible prostate adenocarcinomas and did not involve specific types such as neuroendocrine (< 2%) 37 or intraductal carcinoma (< 1%) 38 , and these special types of tumors often showed lower expression of PSMA, so 68 Ga-PSMA-617 PET/CT examinations might be prone to false-negative results 39,40 that would interfere with their specificity. We acknowledge that further research and analysis are required to address these issues.
There are limitations associated with this study. First, the sample size was low; although participants were selected from among 115 men, 67 patients were included in the final analysis and the number of PCa cases was 33. A larger trial is warranted. Second, the diagnosis of PCa was based on biopsy, and a benign prostate was Scientific RepoRtS | (2020) 10:10963 | https://doi.org/10.1038/s41598-020-67385-9 www.nature.com/scientificreports/ defined on the basis of follow-up findings. Although we conducted these clinical practices according to NCCN guidelines to obtain pathological diagnosis of PCa and benign diagnosis, there were still limitations. For one, the possibilities of missed diagnoses by biopsy persisted, and considering the slow progression of PCa, the follow-up duration may not be enough, leading to the possibility of recurrence or tumorigenesis. Finally, not all the tissue samples underwent immunohistochemical analysis, which might interfere with the accuracy of differentiating histopathological subtypes and the analysis of PSMA expression.
Conclusion. 68 Ga-PSMA-617 PET/CT imaging has better diagnostic performance, especially in terms of specificity, than mpMRI in suspected PCa patients with PSA levels of 4-20 ng/ml. Given the positive correlation between the uptake values of 68 Ga-PSMA-617 PET/CT (SUVmax or SUVratio) and GS, 68 Ga-PSMA-617 PET/ CT could be used as an objective non-invasive imaging tool to predict PCa risk and determine the degree of malignancy. SUVratio seemed to be a more reliable diagnostic parameter of 68 Ga-PSMA-617 PET/CT for detecting PCa with PSA levels of 4-20 ng/ml.
Ethical approval and consent to participate. All procedures performed in studies involving human participants were in accordance with the ethical standards of the institutional and/or national research committee and with the 1964 Helsinki declaration and its later amendments or comparable ethical standards. The study was approved by the Ethics Committee of the Xijing Hospital, Fourth Military Medical University (permit: KY20162088-1). Informed consent was obtained from all participants included in the study.