Visual Abstract
Abstract
18F-DCFPyL, 18F-sodium fluoride (18F-NaF), and 18F-FDG PET/CT were compared in a prospective cohort of men with metastatic prostate cancer (PCa). Methods: Sixty-seven men (group 1) with documented metastatic PCa underwent 18F-DCFPyL and 18F-NaF PET/CT and a subgroup of 30 men (group 2) underwent additional imaging with 18F-FDG PET/CT. The tracers were compared for their detection rates, imaging concordance, associations with prostate-specific antigen (PSA), treatment at the time of imaging, and castration status. Results: Overall, 61 men had metastatic disease detected on one or more scans, and 6 men had no disease uptake on any of the PET/CT scans (and were subsequently excluded from the analysis). In group 1, 18F-NaF detected significantly more metastatic lesions than 18F-DCFPyL (median of 3 lesions vs. 2, P = 0.001) even after eliminating benign causes of 18F-NaF uptake. This difference was particularly clear for men receiving treatment (P = 0.005) or who were castration-resistant (P = 0.014). The median percentage of bone lesions that were concordant on 18F-DCFPyL and 18F-NaF was 50%. In group 2, 18F-DCFPyL detected more lesions than 18F-FDG (median of 5 lesions vs. 2, P = 0.0003), regardless of PSA level, castration status, or treatment. The median percentage of lesions that were concordant on 18F-DCFPyL and 18F-FDG was 22.2%. This percentage was slightly higher for castration-resistant than castration-sensitive men (P = 0.048). Conclusion: 18F-DCFPyL PET/CT is the most versatile of the 3 PET agents for metastatic PCa; however, 18F-NaF detects more bone metastases. Imaging reveals substantial tumor heterogeneity with only 50% concordance between 18F-DCFPyL and 18F-NaF and 22% concordance for 18F-DCFPyL and 18F-FDG. These findings indicate considerable phenotypic differences among metastatic lesions.
Prostate cancer (PCa) is the second leading cause of cancer death among men in the United States, with a 5-y survival rate of only 31% in men with metastatic disease (1). In recent years, precision medicine has offered the hope of improving outcome with treatments tailored to the molecular and clinical characteristics of an individual patient’s malignancy (2,3).
In this context, several targeted radiotracers have emerged to assess PCa by PET/CT. 18F-sodium fluoride (18F-NaF) demonstrates uptake at sites of bone remodeling and osteoblastic activity, with higher sensitivity and specificity for detecting bone metastases than conventional bone scintigraphy (4,5). 18F-DCFPyL targets prostate-specific membrane antigen (PSMA), a membrane glycoprotein highly expressed on PCa cells, especially in metastatic disease (6–8). The most widely used PET agent, 18F-FDG, reflects glucose metabolism commonly upregulated in malignant cells. Although most localized PCa tumors are not 18F-FDG–avid (9), its uptake increases with aggressive and widely metastatic disease (10). Direct comparisons of these agents could cast light on their relative value in men with metastatic PCa.
Therefore, we prospectively compare the performance of 18F-DCFPyL with 18F-NaF and 18F-DCFPyL with 18F-FDG in men with metastatic PCa to understand patterns of overlap and discordance and their potential significance.
MATERIALS AND METHODS
Patient Selection and Study Design
This single-institution open-label prospective, Health Insurance Portability and Accountability Act–compliant study was approved by the institutional review board (NCT03173924) and radiation safety branch. All patients were enrolled after written informed consent was obtained. Eligibility criteria included men with histopathologically confirmed PCa and identifiable metastatic disease on standard-of-care imaging (CT or conventional bone scan). Exclusion criteria included subjects for whom participating would significantly delay standard therapy. There were no exclusion criteria regarding prior or ongoing therapies. Diagnostic and prior treatment history, castration status, and current treatments were recorded after clinical review of medical records.
PET Imaging Protocol
Group 1 subjects underwent 18F-DCFPyL and 18F-NaF PET/CT on separate occasions but within 35 d of each other. 18F-DCFPyL was intravenously injected (mean injected dose, 291.3 MBq [range, 221.4–399.7 MBq]), followed by a head-to-toe PET/CT scan at a mean time of 121.7 ± 7.9 min after injection. 18F-NaF was administered intravenously (mean injected dose, 125.2 MBq [range, 97.9–201.7 MBq]), followed by a head-to-toe PET/CT at a mean time of 63.7 ± 6.0 min after injection.
A subcohort of 30 patients imaged with 18F-DCFPyL also underwent 18F-FDG PET/CT imaging on a separate occasion (group 2) within 33 d of each other. 18F-FDG was administered intravenously (mean injected dose, 377.0 MBq [range, 327.3–433.7 MBq]), with whole-body scanning at a mean time of 61.4 ± 4.6 min after injection.
Imaging was performed on a 3D time-of-flight–mode Discovery MI DR camera (GE Healthcare) with low-dose (120 kV, 60 mAs) CT-based attenuation correction along with random, normalization, dead time and scatter correction.
When technically feasible and after patient consent, a biopsy of at least 1 suggestive lesion identified on imaging was performed within 4–6 mo of scanning.
Imaging Analysis
PET/CT review and analysis was performed using a MIM workstation (version 6.9.2; MIM Software Inc.) by 3 experienced nuclear medicine physicians. Only lesions that were highly suggestive of metastatic or recurrent disease by consensus were included. Indeterminate lesions were excluded from the analysis. In particular, benign causes of increased uptake on 18F-NaF scans were eliminated from the dataset.
SUV, tumor volume (TV), and total lesion uptake (TLU) were reported for every lesion after a semiautomatic segmentation analysis tool for contouring (PET-Edge) was applied. TLU was calculated as the multiplication of SUVmean and TV for each lesion. All values obtained per person were summed to calculate the TLU at the patient level. The total tumor burden (TB) was calculated as the sum of TV from all reported lesions per person.
When the scan showed extensive disease, with lesions too numerous to delineate manually, a semiautomatic software algorithm based on an SUV threshold was used. The pathologic threshold SUV was set at 3 for both 18F-DCFPyL and 18F-FDG and 10 for 18F-NaF. Physiologic uptake and benign and indeterminate lesions were then removed by the readers so that only highly suggestive foci were included in the analysis. In these men, exact lesion number was impossible to count, thus only the TB and the TLU were recorded.
Lesion detection rates and imaging concordance were determined at the patient level and lesion level for the 3 agents. Positive lesions in the same location on different scans were considered concordant regardless of variation in volume or extent. Lesion detection rate and imaging concordance were correlated with PSA, castration status, and treatment at the time of imaging. Men were considered castration-resistant (CRPC) if they had a history of androgen deprivation therapy (ADT) with castrate serum testosterone (<50 ng/dL) plus biochemical or radiologic progression, and were considered castration-sensitive (CSPC) if they never had ADT or if they had a history of ADT but did not fit criteria of CRPC.
Statistical Analysis
18F-DCFPyL, 18F-NaF, and 18F-FDG PET characteristics (number of lesions, TB, and TLU) were correlated to PSA values using Spearman rank correlation. Differences in imaging PET parameters across individual characteristics, such as castration status and treatment at the time of imaging, were evaluated using the Wilcoxon rank-sum test. Comparisons of number of lesions and TB between 18F-NaF and 18F-DCFPyL (for bone lesions only) and between 18F-FDG and 18F-DCFPyL were performed with the paired Wilcoxon test. Lesions were categorized as concordant or discordant across tracers. Concordance between tracers at the patient level was evaluated using Wilcoxon rank-sum and Spearman rank correlation. All tests were 2-sided, and P values < 0.05 were considered significant.
RESULTS
Population
Overall, a total of 67 patients (median age, 67.8 y; age range, 51–84 y) with documented metastatic PCa met criteria for the protocol between June 2017 and February 2020. Six patients were excluded from the analysis because there was no disease uptake on any of the PET/CT scans; therefore only 61 evaluable patients were analyzed. Seven patients (11.5%) had newly diagnosed metastatic PCa and had not received any treatment at the time of imaging. Further specific patient demographics are listed in Table 1. The mean time between 18F-DCFPyL and 18F-NaF and between 18F-DCFPyL and 18F-FDG scans was 7 d (range, 1–35 d) and 8 d (range, 1–33 d), respectively. Patients did not experience adverse events or clinically detected pharmacologic effects after PET scans.
Patient Demographics (n = 61)
Comparison Between 18F-DCFPyL and 18F-NaF (Group 1)
Patient-Based Detection Rate and Concordance Between Radiotracers
All 61 patients had at least 1 pathologic focus consistent with metastatic bone disease on 18F-DCFPyL. The 18F-NaF detection rate was 77.0% for metastatic bone disease.
The median percentage of bone lesions that were concordant between 18F-DCFPyL and 18F-NaF was 50%. The imaging concordance between 18F-NaF and 18F-DCFPyL was independent of castration status, PSA values, treatment at the time of imaging, and time from diagnosis to imaging.
Lesion-Based Detection Rate
A total of 412 bone lesions were detected by 18F-DCFPyL or 18F-NaF. Lesions from 6 patients with extensive disease (“superscans”) were excluded from this analysis because an accurate lesion count was not feasible. 18F-NaF detected 373 of 412 (90.5%) bone lesions and 18F-DCFPyL detected 191 (46.4%). A total of 152 of these bone lesions were concordant between 18F-NaF and 18F-DCFPyL, 39 were detected by 18F-DCFPyL only, and 221 were detected by 18F-NaF only (Fig. 1). The median number of bone lesions detected by 18F-NaF was higher than that by 18F-DCFPyL (P = 0.001) (Fig. 2). Lesion tumor volume detected only by 18F-NaF was significantly lower than that of lesions detected by both 18F-NaF and 18F-DCFPyL (P < 0.05). In this population, 18F-DCFPyL identified 450 soft-tissue lesions (186 pelvic lymph nodes, 112 retroperitoneal lymph nodes, 92 distant lymph nodes, and 11 visceral lesions) in addition to the bone lesions.
Lesion number comparisons.
18F-DCFPyL (left) and 18F-NaF (right) discordance. A 61-y-old man with metastatic PCa involving lymph nodes and bones. PSA at imaging was 49.69 ng/mL. More bone lesions were seen with 18F-NaF than with 18F-DCFPyL (arrows).
Correlation with PSA
The number of lesions, TLU, and total TV derived from 18F-DCFPyL and 18F-NaF correlated with PSA and PSA velocity (Table 2). The strongest correlation was seen between PSA and TLU (ρ = 0.6, P < 0.001) and total TV (ρ = 0.55, P < 0.001) detected by 18F-DCFPyL. These PET metrics showed a weak correlation with PSA doubling time.
Correlation of PSA Characteristics with PET Metrics Derived from 18F-DCFPyL, 18F-NaF, and 18F-FDG, Using Spearman Correlation Coefficient
The median number of bone lesions detected by 18F-NaF was slightly higher than that by 18F-DCFPyL at low PSA levels and rose with increasing PSA (Fig. 3A). The same trend was noted for TV, with a greater TV detected by 18F-NaF than by 18F-DCFPyL, but the difference was not significant.
Number of lesions detected by 18F-DCFPyL (A), 18F-NaF (B), and 18F-FDG (C) according to median PSA (left), treatment at the time of imaging (middle), and castration status (right).
Correlation with Treatment at the Time of Imaging
Men were subdivided into 2 groups according to their treatment at the time of imaging: 1 group consisted of 27 men receiving treatment (mainly ADT and chemotherapy) and the other group consisted of 34 men with no treatment at the time of imaging. Number of lesions, TB, and TLU were higher in the group receiving treatment than in men without treatment (P values ranging from 0.016 to 0.057) (Table 3).
18F-DCFPyL, 18F-NaF, and 18F-FDG Median Number of Lesions, Total Tumor Volume, and Total Lesion Uptake According to Median PSA, Treatment at the Time of Imaging, and Castration Status
For men without treatment, there was no significant difference in median number of bone lesions detected by 18F-NaF versus 18F-DCFPyL, but more bone lesions were detected by 18F-NaF than by 18F-DCFPyL among men receiving treatment (Fig. 3B). Although the difference was not significant, the same pattern was noted for TV with higher bone tumor volume detected by 18F-NaF in comparison to 18F-DCFPyL.
Correlation with Castration Status
Number of lesions, TB, and TLU showed a positive correlation with CRPC status (P values between 0.005 and 0.042) (Table 3).
In CSPC patients, there was no significant difference in median number of bone lesions detected by 18F-NaF versus 18F-DCFPyL, but more bone lesions were detected by 18F-NaF among CRPC patients (9 vs. 5 lesions; P = 0.014) (Fig. 4A). The TB detected by 18F-NaF was higher than that by 18F-DCFPyL for CRPC patients (P = 0.017) and CSPC patients (P = 0.051).
Comparison of median number of lesions detected by 18F-NaF and 18F-DCFPyL (bone lesions only) (A) and by 18F-FDG and 18F-DCFPyL (B) according to median PSA (left), treatment at the time of imaging (middle), and castration status.
Histopathology
A biopsy was performed in 32 patients (52.5%). Five patients had biopsies from 2 different locations. Among the 37 samples, 5 were prostate gland, 6 lymph nodes, 22 bone lesions, and 4 visceral lesions. Most of the samples (94.6%) demonstrated metastases of PCa. Of 22 bone lesions, 18F-NaF demonstrated 2 false-positives (rib, sacrum) and 20 true-positives. 18F-DCFPyL revealed 2 false-positives (rib, sacrum), 1 false-negative (sternum), and 34 true-positives (19 in bone).
Comparison Between 18F-DCFPyL and 18F-FDG (Group 2)
Patient-Based Detection Rate and Concordance Between Radiotracers
A cohort of 30 patients underwent both 18F-DCFPyL and 18F-FDG PET/CT imaging. The 18F-FDG detection rate was 93.3% on a per-patient basis. The median percentage of lesions that were concordant between 18F-DCFPyL and 18F-FDG was 22% (Fig. 5). Imaging concordance between 18F-DCFPyL and 18F-FDG was higher in men with CRPC (66.5%) than CSPC (20%) (P = 0.019) and was independent of other factors.
Concordant PET metastases. A 64-y-old man with metastatic CRPC. PSA at imaging was 464 ng/mL. Concordant pathologic foci were noted on all scans at T8 vertebral body and left iliac bone, consistent with metastasis.
Lesion-Based Detection Rate
Among the 322 lesions detected by 18F-FDG or 18F-DCFPyL (244 soft-tissue lesions and 78 bone lesions), 68 were concordant, 232 were detected by 18F-DCFPyL only, and 22 were detected by 18F-FDG only. The median number of lesions detected by 18F-DCFPyL was 5 (interquartile range, 3–15.5), which was significantly higher than 18F-FDG (median of 2 lesions [interquartile range, 1–3.5], P = 0.0003).
Correlation with PSA, Treatment at the Time of Imaging, and Castration Status
Most metrics derived from 18F-FDG correlated with PSA and PSA velocity, castration status, and treatment at the time of imaging (Table 2; Fig. 3C).
18F-DCFPyL demonstrated more lesions than 18F-FDG regardless of PSA, treatment, and castration status (Fig. 4B).
The total TV detected by 18F-DCFPyL was greater than that by 18F-FDG in the group with a PSA > 10 ng/mL (P = 0.033), when patients were not on treatment (P = 0.044) and in the CSPC group (P = 0.017).
Histopathology
A biopsy was performed in 17 men who underwent 18F-FDG, revealing 3 false-negatives (iliac, ischium, and prostate) and 15 true-positives for the PET tracer.
DISCUSSION
Accurate assessment of disease burden is essential for the management of patients with metastatic PCa. However, it is unlikely that a single targeted imaging agent will detect all lesions given the heterogeneous nature of metastatic disease (11). With very different mechanisms of radiotracer uptake, the low percentage of concordant lesions among the 3 PET agents studied (18F-DCFPyL, 18F-NaF, and 18F-FDG) supports the concept that many phenotypes of metastases exist, even within the same person.
In this study, 18F-NaF showed the highest sensitivity for bone metastases. These results support the results in the study from Harmon et al. in which bone lesion detection rates for 18F-NaF and a first-generation PSMA-targeting agent were 98.4% and 45.4%, respectively, which are similar to our detection rates (93% for 18F-NaF vs. 46% for 18F-DCFPyL) (12). Our findings also agree with the study by Uprimny et al. in which 18F-NaF PET detected a higher number of metastatic bone lesions than 68Ga-PSMA-11 PET (13). These results differ from 2 other studies that found no difference in diagnostic sensitivity for bone metastases between these 2 radiotracers (14,15).
It has been argued that 18F-NaF scans are susceptible to false-positives due to benign disease mimicking metastases (16). However, in this series, in which histologic confirmation was available in several cases, there were only 2 false-positives among 22 osseous lesions detected with 18F-NaF after trained nuclear medicine physicians eliminated obvious benign lesions from consideration. Because PCa cells induce bone formation in adjacent osteocytes, it is likely that only a few cancer cells can affect many regional osteocytes, leading to an amplification of signal on 18F-NaF scans, heightening sensitivity compared with 18F-DCFPyl. We believe that 18F-NaF reflects active disease but may recognize disease below the detection threshold of 18F-DCFPyL (17,18). The relatively high rates of recurrent disease after 177Lu-PSMA therapy in sites not previously identified suggest there is a reservoir of PSMA-negative metastases in the bone that may be detectable by 18F-NaF but not by PSMA radiotracers (19).
One explanation for the lesion mismatch between 18F-NaF and 18F-DCFPyl is that castration resistance could disproportionately influence the performance of 18F-NaF relative to 18F-DCFPyL (20). In CSPC patients, there was no difference in the median number of bone lesions detected by 18F-NaF versus 18F-DCFPyL, but more bone lesions were detected by 18F-NaF among more heavily pretreated CRPC patients (P = 0.014). In immunohistochemistry studies, only 44% of bone metastases expressed PSMA, and osseous lesions with low PSMA detection were associated with CRPC, which could readily explain our findings of discordance with 18F-NaF (7).
18F-FDG– and PSMA-targeting agents showed low concordance in our study. PSMA-negative, FDG-positive lesions are thought to be more aggressive and are linked with poor outcomes as they are encountered more frequently in amphicrine and neuroendocrine phenotypes of CRPC (21). In our research, 18F-DCFPyL detected significantly more lesions than 18F-FDG (P < 0.0001) on both a per-patient and per-lesion basis regardless of castration or treatment status. In about 10% of men, some lesions were positive on 18F-FDG and negative on 18F-DCFPyL despite an overall higher lesion number seen by18F-DCFPyL, implying that a limited number of metastases may exhibit aggressive metabolic features with low PSMA (FDG+, PSMA–) earlier in the course of disease (22). Indeed, similar to the study by Wang et al. (23), we noted discordance between the 2 scans with 22 of 322 lesions (6.8%) detected by 18F-FDG alone in 8 of 30 patients (27%), of which 3 were CRPC and 5 were CSPC. These lesions may be clinically relevant, as decreased survival and therapeutic response have been noted in men with abnormal 18F-FDG PET findings (21,22,24,25). 18F-FDG uptake has been suggested as a biomarker for CRPC and when accompanied by negative 18F-DCFPyL findings, may suggest evolution to neuroendocrine prostate cancer (26). Interestingly, as the disease progressed from CSPC to CRPC, concordance between 18F-FDG and 18F-DCFPyL scans increased (P = 0.048). The discordance among 18F-DCFPyL, 18F-FDG, and 18F-NaF scans in individual lesions confirms phenotypic heterogeneity of PCa metastases, explaining, in part, the difficulty in eradicating such lesions.
The main limitation of this study was the lack of histologic proof for many of the suspected metastases. However, where biopsies were obtained, they overwhelmingly confirmed the presence of cancer in positive scans. Furthermore, readers had access to PET/CT images obtained with the other radiotracers, which may have biased the interpretation of faint uptake when scans were evaluated. However, these unmasked readings reflect daily practice. Finally, the metastatic population was broadly diverse and further investigation stratified by prior therapy may help clarify the respective roles of these radiotracers in the various states of PCa.
CONCLUSION
Imaging men with metastatic PCa using 18F-NaF, 18F-DCFPyL, and 18F-FDG PET demonstrated that 18F-DCFPyL had the best overall performance, but concordance with other agents was low, reflecting phenotypic tumor differences. 18F-NaF identified a significantly higher number of metastatic bone lesions than 18F-DCFPyL. Our study suggests that 18F-NaF might provide additional staging information compared with 18F-DCFPyL, especially in castration-resistant patients and patients receiving treatment at the time of imaging. 18F-DCFPyL functioned better than 18F-FDG in overall lesion detection and was more concordant in CRPC. Further research is warranted to elucidate the utility of 18F-FDG PET and 18F-NaF as prognostic tools and complementary agents to 18F-DCFPyL in understanding tumor heterogeneity patterns in PCa metastases.
DISCLOSURE
This project has been funded in whole or in part with federal funds from the National Cancer Institute, National Institutes of Health, under contract nos. 75N91019D00024, Task Order no. 75N91019F00129, and HHSN261200800001E. The content of this publication does not necessarily reflect the views of policies of the Department of Health and Human Services, nor does mention of trade names, commercial products, or organization imply endorsement by the US. Government. Aloÿse Fourquet is the recipient of a grant from the ARC Foundation for cancer research. No other potential conflict of interest relevant to this article was reported.
KEY POINTS
QUESTION: How does 18F-DCFPyL uptake compare with that of 18F-NaF and 18F-FDG PET/CT in men with metastatic PCa?
PERTINENT FINDINGS: In a prospective study of 67 men with metastatic PCa, 18F-DCFPyL was the most versatile PET agent but 18F-NaF detected more bone metastasis. Substantial tumor heterogeneity was revealed, with only 50% concordance between 18F-DCFPyL and 18F-NaF and 22% concordance between 18F-DCFPyL and18F-FDG.
IMPLICATIONS FOR PATIENT CARE: 118F-FDG and 18F-NaF could be complementary agents to 18F-DCFPyL in staging and illustrating heterogeneous disease characteristics that could optimize treatment strategies for men with metastatic PCa.
ACKNOWLEDGMENTS
We thank Anita Ton, Yolanda McKinney, Juanita Weaver, Philip Eclarinal, Alicia Forest, Chris Leyson, and Mona Cedo for their constant patient care. Thank you to the patients and their families for the sacrifices they made in contributing to this research.
Footnotes
Published online Sep. 2, 2021.
- © 2022 by the Society of Nuclear Medicine and Molecular Imaging.
REFERENCES
- Received for publication March 30, 2021.
- Revision received August 5, 2021.
