Visual Abstract
Abstract
Multiparametric MRI (mpMRI) and prostate-specific membrane antigen (PSMA) PET/CT are complementary imaging modalities used in the presurgical evaluation of patients with prostate cancer (PCa). The purpose of this study was to characterize clinically significant PCa (csPCa) detected and not detected by PSMA PET/CT and mpMRI, focusing on tumors detected solely by PSMA PET/CT and overlooked by mpMRI. Methods: We conducted a single-center, retrospective analysis of patients who underwent both PSMA PET/CT and mpMRI within 3 mo of each other and before radical prostatectomy. Two nuclear medicine physicians and 2 radiologists, in a masked manner, independently contoured PCa lesions on PSMA PET/CT and mpMRI, respectively. A consensus read was done with a third reader for each modality, and a majority rule was applied (2:1). After centralized imaging, a pathologic review was done by a genitourinary pathologist. We assessed agreement between imaging modalities and correlation with pathology. Logistic regression models explored associations between clinicopathologic variables and tumor detection on imaging. Results: In total, 132 csPCa tumors from 100 patients were identified on surgical pathology. PSMA PET/CT showed higher lesion-level (87% vs. 80%) and patient-level (98% vs. 94%) sensitivity than mpMRI. Tumors detected on both imaging modalities were larger and had higher grade groups than those not detected by one or both imaging modalities. On multivariable analysis, csPCa tumors undetected by mpMRI but detected by PSMA PET/CT were smaller than those detected by both modalities. Most tumors showing aggressive pathologic features, such as the large cribriform pattern (94.7%) and the intraductal carcinoma (96%), were correctly detected by both imaging modalities. Limitations included selection bias in a surgical cohort. Conclusion: PSMA PET/CT tends to detect smaller csPCa not detected by mpMRI. Larger tumors on pathology with higher grade groups are more likely to be correctly detected by both imaging modalities. These findings provide insights for refining presurgical evaluation strategies in PCa.
In the past decade, the advancements and widespread adoption of MRI and the introduction of prostate-specific membrane antigen (PSMA) PET/CT have revolutionized the management of patients with prostate cancer (PCa). PSMA PET/CT has shown its value in the staging of intermediate- and high-risk PCa (1,2), localizing biochemically recurrent PCa (3,4) and guiding definitive and salvage therapies (5,6). Multiparametric MRI (mpMRI) has become the gold standard imaging technique to stage and localize PCa foci (7). mpMRI-targeted biopsies have greatly improved the identification of clinically significant PCa (csPCa) and allowed for the tracking of biopsy cores over time (8,9). However, approximately a third of csPCa is not visible on mpMRI (10).
A recent systematic review of 31 studies demonstrated that although PSMA PET/CT was not superior to mpMRI in terms of T staging, the addition of PSMA PET/CT improved the sensitivity for detecting extracapsular extension from 53% to 79% and the sensitivity for detecting seminal vesicle invasion from 51% to 67% (11). In this context, we hypothesized that csPCa tumors detected by PET and not detected by MRI differed from those detected by both imaging modalities in terms of clinicopathologic variables. We further tested the hypothesis that before surgical imaging, clinical and pathologic variables are associated with whether lesions are true-positive or false-negative on imaging. To perform this test, we performed a retrospective assessment of a prospectively maintained imaging and pathologic dataset. Centralized imaging and pathologic review resulted in a large cohort of patients who underwent prostatectomy with both mpMRI and PSMA PET/CT before surgery.
MATERIALS AND METHODS
Study Design and Patients
This retrospective, single-center analysis received institutional review board approval (approval 10-001863). To identify our cohort, we leveraged the UCLA Integrated Diagnostics database, a prospective dataset that captures detailed clinical, imaging, and pathologic data on patients diagnosed and treated for PCa. Our inclusion criteria were radical prostatectomy at UCLA, whole-mount histopathology, and presurgical PSMA PET/CT and mpMRI within 3 mo of each other. Excluded from the primary analysis were patients who received any neoadjuvant hormonal therapy (Figs. 1 and 2).
Study design flowchart. NM = nuclear medicine.
Patient selection flowchart.
mpMRI Acquisition
mpMRI was performed using one of several 3-T MRI systems (Vida, Skyra, or Prisma; Siemens Healthineers) and a standardized protocol including 2- and 3-dimensional T2-weighted, echoplanar, diffusion-weighted imaging with 3 b-values, as well as 3-dimensional perfusion imaging after intravenous gadolinium injection with pelvic external phased-array coils (12). A small–field-of-view 3-dimensional axial turbo spin-echo T2-weighted sequence was performed using spatial and chemical-shift–encoded excitation (Siemens Healthineers), as described in detail previously (12,13).
PSMA PET/CT Acquisition
PSMA PET/CT images were acquired after intravenous injection of 68Ga-PSMA-11 using a Biograph 64 or mCT PET/CT scanner (Siemens Medical Systems) and a standard protocol (14). Oral and intravenous CT contrast material was administered unless contraindicated. A 5-mm slice thickness was used for the CT images. All PET images acquired from pelvis to vertex were corrected for attenuation, dead time, random events, and scatter.
Centralized Image Analysis
PSMA PET/CT and mpMRI were read independently by two board-certified nuclear medicine physicians and two abdominal fellowship–trained radiologists, respectively. Readers were aware of biopsy-proven PCa but not of clinical information or the findings of the other imaging modality. All scans were centrally anonymized and assigned an individual study number. A nuclear medicine physician and senior abdominal imaging radiologist acted as a third reader in cases of disagreement. Focal uptake on PSMA PET/CT was considered suggestive of PCa on the basis of the PSMA uptake intensity, pattern, and location on a visual assessment. Suspected lesions were manually contoured, and semiquantitative measures, such as SUVmax and SUVmean, were extracted from the PSMA PET/CT lesions. In addition to the manual contouring, an additional volume of interest was defined by including a region with uptake greater than 50% of the SUVmax within the manually drawn volume of interest to enhance standardization. The prostate volume was measured on mpMRI, and a Prostate Imaging–Reporting and Data System (version 2.1) score was documented for each lesion (15). Only lesions with scores of 3–5 were considered positive for the analysis.
Centralized Histopathologic Analysis
Prostate glands were sectioned in the axial plane from base to apex in approximately 5-mm slices using a 3-dimensional printed mold and embedded into paraffin. The tissue was cut into 5-μm sections onto whole-mount slides and stained with hematoxylin and eosin. As part of a clinical read, a genitourinary pathologist evaluated tumor histology, assigned International Society of Urologic Pathology (ISUP) grade groups (GGs), and localized and measured all PCa foci by drawing their contours on the whole-mount slides. GG 1 lesions were considered non-csPCa.
Imaging and Pathologic Correlation
Using a custom-made fusion workflow, we fused the PSMA PET/CT and mpMRI. A nuclear medicine physician and 2 radiologists, working in consensus, assessed the correspondence of the imaging-identified lesions with tumors seen on histopathology, with a genitourinary pathology joining the consensus read when the assessment was considered challenging.
Histopathologically described csPCa corresponding to lesions identified on imaging was considered true-positive; that not detected on imaging was considered false-negative. True positivity was considered to be cases of total and near-total agreement (16). Lesions identified on imaging but not corresponding to csPCa on histopathology were considered false-positive.
Statistical Analysis
Descriptive statistics are expressed as mean ± SD or median and interquartile range. We compared lesion size on pathology and GG and PSMA PET/CT quantitative measures between detected and undetected lesions on imaging. We assessed the diagnostic performance and detection rates of PSMA PET/CT and mpMRI for csPCa defined as having a GG of 2 or higher. We obtained a sensitivity and positive predictive value analysis and accounted for histopathologic lesions corresponding to multiple imaging-identified lesions, and vice versa. Sensitivity was also determined at the patient level. We used independent-samples t testing for continuous variables, Mann–Whitney U testing for ordinal variables, and 1-way ANOVA to assess group differences using the Bonferroni post hoc test for multiple comparisons.
To test our primary hypothesis, we used multivariable logistic regression models to assess the association of clinical, pathologic, and imaging variables with PSMA PET/CT and mpMRI disagreement with each other and with pathology. In the logistic regression model analyses, patient age at time of imaging and prostate volume measured on mpMRI were treated as continuous variables, with increments of 10 y and 10 cm3, respectively. GG was treated as a categoric value with 3 groups (GG 2, GG 3, and GG 4/5). Serum prostate-specific antigen levels and lesion size on pathology were treated as continuous variables per increase of 1 ng/mL and 1 cm, respectively.
We additionally replicated all assessments in the primary cohort, including the patients who were excluded because they received neoadjuvant hormonal therapy. Statistical analyses were conducted using SPSS version 28 (IBM) and the DATAtab online statistics calculator. P values of less than 0.05 were considered statistically significant.
RESULTS
Patients
The final cohort comprised 114 patients, 14 of whom were excluded from the primary analysis because they had received neoadjuvant androgen deprivation therapy (primary cohort, n = 100). Figure 2 shows the patient selection flowchart. Patient and tumor characteristics are summarized in Table 1. The median serum prostate-specific antigen level was 8.6 ng/mL (interquartile range, 5.1–15.2), and most patients had presurgical unfavorable intermediate-risk (30/100, 30%) or high-risk (53/100, 53%) PCa based on the National Comprehensive Cancer Network risk classification.
Patient and Tumor Characteristics
PCa Detection by MRI or PET Alone
In total, 156 tumors were identified on histopathology, and 132 (85%) were considered csPCa (Table 1). Most patients had 1 csPCa tumor (76%), whereas fewer had 2 (19%) or 3 (5%) tumors. PSMA PET/CT and mpMRI identified 143 and 122 lesions, respectively. Nineteen of 143 (13%) and 8 of 122 (7%) lesions were false-positive on PSMA PET/CT and mpMRI, respectively. Among the false-positive findings, both PSMA PET/CT and mpMRI identified 3 lesions corresponding to GG 1 tumors (Table 2). PSMA PET/CT and mpMRI detected 115 of 132 (87%) and 106 of 132 (80%) tumors, respectively (Table 2). PSMA PET/CT and mpMRI did not detect 17 of 132 (12%) and 26 of 132 (20%) csPCa tumors, respectively. Sensitivity for the identification of csPCa was 87% for PSMA PET/CT and 80% for mpMRI, and the positive predictive value for PSMA PET/CT and mpMRI was 87% and 93%, respectively (Table 2; Supplemental Table 1; supplemental materials are available at http://jnm.snmjournals.org). Sensitivity at the patient level was 98% and 94% for PSMA PET/CT and mpMRI, respectively.
Sensitivity and Agreement Analysis for PSMA PET/CT and mpMRI
Tumor size was larger and GGs were higher among true-positive than false-negative lesions on both PSMA PET/CT and mpMRI (Fig. 3; Table 3). A large cribriform pattern (LCP) and intraductal carcinoma (IDC) on pathology were associated with positive detection by both imaging modalities (Table 3).
Distribution of ISUP GG and size on histopathology based on imaging findings. aP values determined from t test for independent samples. bP values determined from Mann–Whitney U test. cBonferroni post hoc tests. ISUP = International Society of Urological Pathology.
Patient and Clinically Significant Tumor Characteristics
Multivariable logistic regressions showed that smaller tumors on pathology (odds ratio, 0.24; 95% CI, 0.10–0.58) and older age (odds ratio, 7.57; 95% CI, 1.54–37.1) were associated with nondetection of csPCa by PSMA PET/CT (Table 4). A smaller tumor size on pathology (odds ratio, 0.21; 95% CI, 0.10–0.46) was associated with nondetection of csPCa by mpMRI, whereas GG 4/5 versus GG 2 (0.09; 95% CI: 0.01–0.66) was associated with lower odds of nondetection (Table 4).
Multivariable Logistic Regression for Clinicopathologic Variables Associated with Detected vs. Nondetected Lesions and for PET+/MRI− Lesions vs. PET+/MRI+ Lesions
Imaging Concordance
On whole-mount histopathology, csPCa tumors on pathology were divided into 4 agreement groups based on the imaging findings: PET+/MRI+, PET−/MRI−, PET+/MRI−, and PET−/MRI+. PSMA PET/CT and mpMRI both correctly identified 101 of 132 lesions (77%), and neither detected 12 of 132 csPCa lesions (9%). PSMA PET/CT and mpMRI correctly detected 14 of 132 (11%) and 5 of 132 (4%), respectively, csPCa tumors not detected by the other imaging modality (Table 5).
Patients and Clinically Significant Tumor Characteristics Stratified by PSMA PET/CT and mpMRI Agreement
Among the clinicopathologic and imaging variables tested, age, GG, tumor size on pathology, LCP, and IDC significantly differed among the 4 agreement groups (Table 5; Fig. 3). We sought to characterize tumors detected by PET and not mpMRI by comparing the agreement groups PET+/MRI+ and PET+/MRI− using multivariable logistic regressions. Only decreasing tumor size on pathology was associated with PET+/MRI− (Table 4). The same analysis in MRI+ csPCa not detected by PSMA PET/CT was not conducted because of the small sample size (PET−/MRI+, n = 5/132).
Most tumors with LCP on pathology (71/75, 94.7%) were correctly detected, whereas 1 of 75 (1.3%) was not detected by both imaging modalities, and 1 of 75 (1.3%) and 2 of 75 (2.7%) were not detected by PSMA PET/CT only and mpMRI only, respectively. Most tumors showing IDC on histopathology (48/50, 96%) were correctly identified by both imaging modalities, and only 2 of 50 tumors (4%) were not detected by mpMRI only (Table 5).
Subanalysis Including Patients Given Neoadjuvant Androgen Deprivation Therapy
Compared with results from the primary analyses, findings were only minimally changed when patients who were treated with neoadjuvant androgen deprivation therapy were included (Supplemental Tables 2 and 3).
DISCUSSION
Although existing literature from mpMRI studies has established that large, solitary, aggressive tumors are more likely to be visualized, approximately a third of primary csPCa is overlooked by mpMRI (10,17). PSMA PET/CT can play a role in complementing this lack of sensitivity of mpMRI. This single-center analysis leveraged a large cohort of patients who underwent radical prostatectomy after both mpMRI and PSMA PET/CT, with central imaging and pathologic assessment to compare the 2 imaging modalities within patients. The objectives of our analysis were to assess differences in detection rates of csPCa, characterize the tumors detected and not detected by each imaging modality, and identify the clinicopathologic and imaging variables associated with tumors detected by PSMA PET/CT and not detected by mpMRI.
According to a recently published metaanalysis comparing PSMA PET/CT and mpMRI in detecting csPCa, PSMA PET/CT has higher sensitivity than mpMRI (87% vs. 84%) while maintaining comparable diagnostic accuracy (18). Other studies showed combined PSMA PET/CT and mpMRI to improve accuracy for the initial local (T and N) staging for PCa (19). The results of our analysis, based on masked reads of the presurgical imaging scans, revealed that PSMA PET/CT has higher lesion-level (87% vs. 80%) and patient-level (98% vs. 94%) sensitivity but a lower positive predictive value than mpMRI (87% vs 93%), confirming the complementarity of the 2 imaging modalities. However, whether these results justify the widespread combined use of PSMA PET/CT and mpMRI, or the use of PSMA PET/MRI, still needs to be understood.
To address this lack of knowledge, our study investigated the discrepancies between the 2 imaging modalities and final pathology. Although a study by Johnson et al. showed that larger tumors with higher grade are more likely detected by mpMRI (10), the characteristics of lesions detected and undetected on PSMA PET/CT were not previously assessed. Our results highlight that tumors showing a higher GG and a larger size on pathology were more likely to be detected on both imaging modalities. A younger age was also found to be related to a higher likelihood of csPCa detection on PSMA PET/CT but not mpMRI. The reasons behind this association are not clear and could be related to a small number of events (n = 17) or underlying confounding factors between characteristics. This finding would need to be further explored in a future validation study. In addition, detecting LCP and IDC before a treatment decision can also be useful given that they are associated with more aggressive disease (20,21). Both PSMA PET/CT and mpMRI detected most tumors showing LCP (94.7%) or IDC (96%). Only 4% of tumors with LCP or IDC were not detected by one of the two imaging modalities, whereas 1.3% of tumors with LCP and none of those with IDC were missed by both imaging methods. Overall, larger tumors showing more aggressive pathologic patterns were more likely to be detected by both PSMA PET/CT and mpMRI.
The characteristics of MRI-occult csPCa detected by PSMA PET/CT have never, to our knowledge, been investigated before at the individual level. Therefore, the role of PSMA PET/CT before definitive management for localized disease remains largely unknown. In this study, we showed that incorporating PSMA PET/CT with mpMRI identifies more csPCa tumors invisible on mpMRI. In this cohort, PSMA PET/CT detected 14 of 132 csPCa tumors (11%) that were not seen on mpMRI (PET+/MRI−). These tumors tended to be smaller than those seen on both PSMA PET/CT and mpMRI (PET+/MRI+) based on multivariable analyses. This observation aligns with a well-known limitation of imaging modalities—the inability to visualize microscopic disease. Although the overall higher sensitivity of PSMA PET/CT is well established and suggests a possible role in detecting additional intraprostatic tumors (11,22,23), the characterization of these tumors is unique to our study. Only a small percentage of csPCa (5/132, 4%) was not detected by PSMA PET/CT but was detected by mpMRI (PET−/MRI+). This limited our ability to conduct a reciprocal analysis—determining which cases are more likely to be undetected by PSMA PET/CT and detected by mpMRI.
The effects of hormonal treatments on PSMA expression in PCa cells are still poorly understood. Although preclinical and clinical studies indicate an influence of hormone therapy on PSMA expression (24,25), efforts to prospectively assess the effects of short- and long-term hormonal treatments on PSMA expression measured by PET have produced inconclusive results (26–28), ultimately suggesting multifactorial interactions. Hormone therapies have demonstrated effects on tumor visualization and MRI quantitative parameters such as a reduction in prostate and tumor volume, particularly in the peripheral zone; decreased signal intensity on T2-weighted images; and decreased transfer constant value, with an increase in the apparent diffusion coefficient (29–31). The clinical implications of these changes have not been clearly assessed yet. Our sensitivity analysis conducted on a cohort including 14 additional patients who underwent neoadjuvant hormonal therapy before imaging showed no significant differences from the main analysis. This lack of significance may be attributed to the small number of additional patients or the negligible impact of hormonal treatments on PCa detection. Addressing these uncertainties requires further confirmatory studies.
The main limitation of our study was the patient selection bias. Our cohort comprised patients who underwent radical prostatectomy for biopsy-proven PCa. The imaging readers, despite being masked to any clinical information, were aware that patients had PCa, potentially limiting the value of the positive predictive value, which should be interpreted with caution in this selected population. Additionally, our cohort was predominantly patients with unfavorable intermediate- and high-risk PCa. Although this does not represent the full spectrum of PCa, it reflects the population more likely to undergo both PSMA PET/CT and mpMRI in a clinical setting. Another limitation is the nonhomogeneous interval between PSMA PET/CT and mpMRI because of the retrospective nature of this study. To address this limitation, we selected a relatively restrictive interval of 3 mo, though we acknowledge that this still may allow for minor tumor changes over time. Additionally, the lack of clinical follow-up did not allow us to clearly discern the value PSMA PET/CT adds to mpMRI in clinical decision making or prognostics.
CONCLUSION
PSMA PET/CT has higher sensitivity than mpMRI for detecting csPCa. Tumors detected by PSMA PET/CT and invisible on mpMRI tend to be smaller and have lower GGs than mpMRI-visible tumors. A higher GG and larger tumor size on pathology are found in tumors correctly identified by both imaging modalities. Our findings highlight the potential value of combining PSMA PET/CT with mpMRI by elucidating the specific clinical, pathologic, and imaging characteristics of mpMRI-occult lesions that are identified on PSMA PET/CT. Prospective studies are needed to assess the value PSMA PET/CT adds to mpMRI in clinical decision making or prognostics.
DISCLOSURE
This work was funded by the Department of Radiological Sciences, David Geffen School of Medicine at UCLA. Data for this analysis were provided by the Integrated Diagnostics Shared Resource, a collaboration between the Departments of Radiological Sciences, Pathology, and Laboratory Medicine, and Molecular and Medical Pharmacology–Nuclear Medicine Division. Johannes Czernin is a founder of SOFIE Biosciences and holds equity in the company and in intellectual property invented by him, patented by the University of California, and licensed to SOFIE Biosciences. He is a founder and board member of Trethera Therapeutics and holds equity in the company and in intellectual property invented by him, patented by the University of California, and licensed to Triangle. He serves on the medical advisory board of Actinium Pharmaceuticals and on the scientific advisory boards of POINT Biopharma, RayzeBio, and Aktis Oncology. Jeremie Calais reports prior consulting activities outside the submitted work for Advanced Accelerator Applications, Blue Earth Diagnostics, Curium Pharma, GE HealthCare, Janssen, POINT Biopharma, Progenics, Radiomedix, and Telix Pharmaceuticals. He is the recipient of grants from the Prostate Cancer Foundation (2020 Young Investigator Award, 20YOUN05), the Society of Nuclear Medicine and Molecular Imaging (2019 Molecular Imaging Research Grant for Junior Academic Faculty), the Philippe Foundation Inc. (NY, USA), and the ARC Foundation (France) (International Mobility Award SAE20160604150). Adam Weiner was supported by the Simon-Strauss Foundation, the UCLA Dr. Allen and Charlotte Ginsburg Fellowship in Precision Genomic Medicine, and the Prostate Cancer Foundation Young Investigator Award (23YOUN21). No other potential conflict of interest relevant to this article was reported.
KEY POINTS
QUESTION: What clinical, pathologic, and imaging parameters are associated with PCa detection by PSMA PET/CT and mpMRI?
PERTINENT FINDINGS: Tumors detected by PSMA PET/CT and invisible on mpMRI tend to be smaller and have a lower ISUP GG than mpMRI-visible tumors. Higher ISUP GG, larger tumor size, and aggressive pathologic features, such as LCP and IDC, are found in tumors correctly identified by both imaging modalities.
IMPLICATIONS FOR PATIENT CARE: The use of both PSMA PET/CT and mpMRI allows the identification of more csPCa; specifically, the addition of PSMA PET/CT is useful for the visualization of smaller tumors with lower ISUP GGs.
Footnotes
Published online Oct. 30, 2024.
- © 2024 by the Society of Nuclear Medicine and Molecular Imaging.
REFERENCES
- Received for publication July 16, 2024.
- Accepted for publication September 30, 2024.
In this issue
Jump to section
Related Articles
Cited By...
- No citing articles found.