Abstract
Accurate detection of prostate cancer lymph node metastases (LNM) through PET/CT before lymphadenectomy is crucial for successful therapy. PET/CT with choline derivatives used to be the standard tool for imaging metastases, whereas 68Ga-PSMA (prostate-specific membrane antigen) PET/CT was introduced recently. Both PET techniques were investigated with respect to what extent the detection rate of LNM depends on the size of tumor deposits (TDs) within LNM. Methods: Documenting the switch from the use of 18F-choline to 68Ga-PSMA in 2014, we used 2 patient cohorts undergoing a template lymphadenectomy because of a PET/CT indicating LNM. Forty-four and 40 patients underwent PET/CT with 18F-choline or 68Ga-PSMA ligand, respectively. In total, 226 LNM (125 18F-choline, 101 68Ga-PSMA) originated from 73 salvage lymphadenectomies at biochemical recurrence and from 11 primary lymphadenectomies at radical prostatectomy. LNM eligible for direct correlation of PET/CT to histopathology were identified from lymphadenectomies conducted in small anatomic subregions, with 1 LNM (condition 1) or 1–2 LNM (condition 2). Longitudinal and short diameters of TD within LNM were determined by histopathology, allowing linking of the size of TD in LNM to the detection threshold of PET/CT. Diameters associated with a detection rate of 50% and 90% (d50%, d90%) were calculated on the basis of logistic growth curve models fitted. Results: Gleason score, number of removed LNs, and subregions for lymphadenectomy per patient did not differ significantly between the 18F-choline and 68Ga-PSMA groups. The median prostate-specific antigen level at imaging and number of LNM per patient were significantly higher in the 18F-choline group (3.4 ng/mL, n = 34) than in the 68Ga-PSMA group (2.2 ng/mL, n = 28; both P < 0.05). Longitudinal and short diameters of TD in LNM to reach d90% were 11.2 and 7.4 mm, respectively, for 18F-choline PET/CT and 6.3 and 4.9 mm, respectively, for 68Ga-PSMA PET/CT. Corresponding diameters to reach d50% were 5.5 and 3.3 mm, respectively, for 18F-choline PET/CT and 3.7 and 2.3 mm, respectively, for 68Ga-PSMA PET/CT. Detection rates were significantly higher under 68Ga-PSMA (P = 0.005 and 0.04 for longitudinal and short diameter). Conclusion: 68Ga-PSMA PET/CT is superior to 18F-choline PET/CT in the detection of LNM. Whether those results will lead to an improved patient outcome after 68Ga-PSMA PET–guided therapy needs to be investigated by further studies.
A precise detection of lymph node (LN) metastases (LNM) in prostate cancer (PCa) is fundamental at primary diagnosis and at the stage of biochemical recurrence (BR) with PCa relapse (mainly due to LNM) to plan and conduct the most adequate therapy (1–3). Surgical resection of LNM through pelvic lymphadenectomy associated with radical prostatectomy (RP) in the case of primary therapy or associated with salvage lymph node dissection (salvage lymphadenectomy) in the case of BR are approaches offered to patients with suspected LNM. Those therapies might be able to reduce the tumor burden significantly, delay clinical progression, and avoid early systemic treatment such as hormone deprivation therapy (ADT) or chemotherapy (1,2,4).
As conventional imaging modalities such as CT and MRI exhibit a limited sensitivity for LNM detection (5), PET/CT with choline analogs such as 18F-fluoroethyl-choline or 11C-choline was the standard tool for imaging PCa lesions for many years. PET/CT imaging of the expression of the prostate-specific membrane antigen (PSMA) was introduced only recently (3,5–9). Targeting PSMA on the surface of PCa cells allows for powerful imaging of PCa lesions (8,10,11). The superiority of 68Ga-PSMA PET/CT compared with 18F-choline PET/CT with respect to general diagnostic accuracy has been analyzed by different groups (3,12,13). These studies did not, however, stratify the results by the size of the tumor deposits (TDs) in LNMs (3).
In this present study, we extend these investigations by analyzing to what extent the detection of LNMs depends on the size of the TD in LNM from lymphadenectomy after 18F-choline PET/CT or 68Ga-PSMA PET/CT. Such an investigation requires LNMs for which we can establish a direct correlation from the PET/CT results to histopathology, as the latter evaluates subregions, not single LNMs. A direct correlation is possible, if there is only 1 LNM (verified by histopathology) in 1 subregion rated by PET/CT. This allowed us to compare the detection rates for TD in LNMs of a given size between 18F-choline PET/CT and 68Ga-PSMA PET/CT. Describing the conversion from 18F-choline (2007–2014) to 68Ga-PSMA in 2014, we used 2 available patient cohorts from our clinic undergoing a template lymphadenectomy because of a PET/CT indicating LNM (11,14). From both cohorts, subregions eligible for conditions 1 and 2 were included in this study. Data were collected from 73 patients at the stage of BR undergoing a salvage lymphadenectomy and from 11 men at the stage of primary therapy (RP with extended lymphadenectomy). The detection rate of PET/CT on the size of TD in LNM was investigated in 125 (18F-choline) and 101 LNMs (68Ga-PSMA). Diameters associated with a detection rate of 50% and 90% (d50%, d90%) were calculated on the basis of logistic growth curve models fitted.
MATERIALS AND METHODS
Study Goal and Design
The aim of this retrospective study was to analyze the dependence of the LNM detection rate of 18F-choline and 68Ga-PSMA PET/CT on the size of the TD in LNM from salvage lymphadenectomy and primary lymphadenectomy. Such an investigation requires LNMs for which we can establish a direct correlation from the PET/CT results to histopathology, as the latter evaluates subregions, not single LNMs. A direct correlation is possible, if there is only 1 LNM (verified by histopathology) in 1 subregion. If there are 2 LNMs in 1 subregion and only 1 is detected, we do not know exactly which one was detected, but mathematic modeling still allows us to establish a link. In our study, we therefore considered only subregions with 1 or 2 LNMs, allowing us to compare the detection rates for TD in LNMs of a given size between 18F-choline PET/CT and 68Ga-PSMA PET/CT. Because the analysis of subregions with 1 LNM is easier than the analysis with 2 LNMs, we present results both for subregions with only 1 LNM (condition 1) and for subregions with 1 or 2 LNMs (condition 2) (Supplemental Fig. 1; supplemental materials are available at http://jnm.snmjournals.org).
Source of Patients
We used 1 patient cohort undergoing a template lymphadenectomy because of a PET/CT indicating LNM at our clinic and (published previously (11,14)) supplemented by patients with lymphadenectomy at RP at our center. Because of a shift from 18F-choline to 68Ga-PSMA as the tracer for PET/CT in 2014, this cohort can be divided into 2 corresponding groups. In each group, those patients with at least 1 subregion satisfying condition 1 or condition 2 were selected: 44 individuals in the 18F-choline group originated from the patient cohort published in 2014 by our group (n = 72) (14). Twenty-nine of 40 individuals from the 68Ga-PSMA group originated from the patient cohort published in 2017 by our group (n = 30) (11); the remaining 11 patients underwent an extensive template lymphadenectomy at RP between November 2014 and December 2015 (Table 1; Fig. 1). The institutional review board approved this study (No. 562/15), and all subjects signed a written informed consent form.
Clinical Characteristics from 84 PCa Patients Undergoing Lymphadenectomy
Workflow with surgery, tissue sample processing, and identification of eligible subregions in men with suspected nodal metastatic PCa undergoing a template lymphadenectomy after PET/CT. Tissue specimens from small anatomic subregions were removed separately at surgery. At histopathology, the size of TDs was determined. Aiming at a direct link from PET/CT to histopathology, subregions fulfilling the requirement for conditions 1 and 2 were selected from 18F-choline and the 68Ga-PSMA PET/CT groups. Finally, size of TDs in detected and undetected LNMs was evaluated for estimation of detection thresholds.
PET Tracer
18F-choline PET/CT was the available tracer for imaging PCa patients at BR from 2007 to 2014. In 2014, 68Ga-PSMA PET/CT was introduced at our center and replaced 18F-choline PET/CT. We therefore observed the shift from 18F-choline to 68Ga-PSMA with respect to LNM detection.
LNMs Available
Overall, 84 patients provided subregions with 1 single LNM (condition 1) and 1 or 2 LNMs (condition 2) verified by histopathology (Fig. 1; Table 1; Supplemental Fig. 1). For condition 1, we included 52 detected and 21 undetected single LNMs in the 18F-choline PET/CT group and 55 detected and 18 undetected LNMs in the 68Ga-PSMA PET/CT group; for condition 2, we included 74 detected and 51 undetected LNMs in the 18F-choline PET/CT group and 72 detected and 29 undetected LNMs in the 68Ga-PSMA PET/CT group (Table 1).
18F-Choline-PET/CT and 68Ga-PSMA-HBED-CC PET/CT Imaging Analysis
The 18F-choline PET/CT and 68Ga-PSMA PET/CT were conducted as previously described by Jilg et al. (11,14). PET/CT was performed after intravenous injection of a mean activity ± SD of 245 ± 40.5 MBq of 18F-fluorethylcholine and 189.6 ± 32.6 MBq of 68Ga-PSMA (Glu-NH-CO-NH-Lys-(Ahx)-[68Ga(HBED-CC)]) with image acquisition starting 45–60 min after injection. Transverse PET/CT slices, 3 mm in thickness, were generated. CT data were used to correct for photon attenuation. A 18F-choline– or a PSMA-positive lesion was defined as a focal tracer accumulation greater than normal or physiologic local background activity. Two experienced nuclear medicine physicians evaluated all the PET/CT studies in consensus by side-by-side review of the coregistered PET and CT datasets using predefined PET window settings (inverted gray scale; SUV range, 0–5 g/mL) (PAXCS and IMPAX workstation [AGFA Health Care]). The 2 PET/CT readers were masked to the data from histopathology.
Lymphadenectomy
Salvage Lymphadenectomy
Patients at the stage of BR (PSA > 0.2 ng/mL in 2 consecutive measurements) after primary therapy and the suspicion of LNM (without detectable bone or visceral metastases) in an 18F-choline PET/CT (n = 44) or 68Ga-PSMA PET/CT (n = 29) underwent a salvage lymphadenectomy on a compassionate-use basis. Depending on the presence of PET-positive lesions (pelvic, retroperitoneal, or both), a bilateral template pelvic lymphadenectomy (subregions: common iliac vessels, external iliac vessels, obturator vessels, internal iliac vessels, presacral region), a retroperitoneal lymphadenectomy (subregions: aortic bifurcation, aortal, caval, interaortocaval), or combined pelvic and retroperitoneal lymphadenectomy was conducted.
Primary Lymphadenectomy
Eleven patients underwent a primary extended lymphadenectomy at RP after a 68Ga-PSMA PET/CT with the suspicion of LNMs. Because of the presence of PET-positive pelvic regions, a bilateral template pelvic lymphadenectomy (subregions: common iliac vessels, external iliac vessels, obturator vessels, internal iliac vessels, presacral region) was conducted.
Generally, whenever intraoperative circumstances permitted, we adhered to the templates at salvage lymphadenectomy or primary lymphadenectomy and removed the nodal-fibro-fatty tissue. During lymphadenectomy, LN portions from each subregion were collected separately.
Histopathology
All resected LNs were formalin-fixed and paraffin-embedded, followed by histopathologic evaluation by 1 pathologist on hematoxylin and eosin–stained tissue slides. The pathologist was not aware of the PET findings nor was he informed about the clinical results provided by the surgeon. The longitudinal diameter and short diameter of each whole LNM and TDs were measured under the microscope, in millimeters, using a ruler. The short diameter represents the maximum diameter we could observe when looking orthogonal in the direction of the longitudinal diameter. Area and volume of the TD in LNM were determined assuming an ellipsoid shape and hence the following formulas were used:
Statistics
The outcome of interest was the detection of an LNM, that is, a true-positive result of PET/CT. In a first step, we divided the LNMs identified under condition 1 according to their longitudinal and short diameter in intervals of 1 mm in length and considered the empiric detection rate in each interval, that is, the proportion of LNMs detected by PET/CT. We fitted logistic growth curve models, describing the probability (on the logit scale) of detecting an LNM as a linear function 
α + βd 
of the diameter d. From these models, we estimated the diameters d50% and d90% associated with a detection rate of 90% and 50%, respectively. Fitting such models is also possible under condition 2, allowing the relation of 2 diameters to the observed PET/CT results; details are given in the Supplemental Appendix. The statistical significance of the overall trend toward smaller or greater detection rates in 1 group was assessed by fitting a joint growth curve model so as to allow only a shift between the 2 curves. To judge the clinical relevance of the estimated diameters associated with a certain detection rate, we also determined the relative frequency of LNMs with diameters exceeding the estimated ones. These numbers estimate the actually observed fractions of LNMs with detection probabilities above 50% or 90% in our clinical sample, respectively. Statistical computations are based on Stata 14.2 (StataCorp [https://www.stata.com/]) and Prism7 software (GraphPad).
RESULTS
Clinical Parameters and Results from Lymphadenectomy
Age at surgery, Gleason score, time from PET/CT to surgery, number of removed LNs and LNMs, subregions for lymphadenectomy per patient, and number of removed LNs per subregion did not differ between the individuals who underwent 18F-choline PET/CT or 68Ga-PSMA PET/CT. In contrast, men from the 18F-choline PET/CT cohort exhibited a higher PSA level and a higher number of LNMs than did those in the 68Ga-PSMA PET/CT group (Table 1). Figure 1 shows the workflow concerning patient selection, surgery, histopathologic workup, and our approach for the evaluation of the detection rates (conditions 1 and 2). Four representative examples (PET with corresponding CT) of LNM with direct link from PET to histopathology on a 18F-choline PET/CT or 68Ga-PSMA PET/CT are shown in Supplemental Figures 2A and 2B. Clinical parameters and data from lymphadenectomy are listed in Table 1. Information about the number of LNMs for conditions 1 and 2 for both tracer groups are shown in Table 2.
Number of LNs Included for Conditions 1 and 2 from 18F-Choline and 68Ga-PSMA PET/CT Groups
Characteristics of LNMs
The distribution of the longitudinal diameter and short diameter and area and volume of TDs in detected or overlooked LNMs under condition 1 (overall n = 146 LNMs) by 18F-choline PET/CT (n = 73 subregions, LNM) or 68Ga-PSMA PET/CT (n = 73 subregions, LNM) are described in Supplemental Table 1.
Detection Rates
The relation of detection rates for 18F-choline PET/CT or 68Ga-PSMA PET/CT to diameters is visualized in Figures 2 and 3, including visualizations of the fitted growth curves and the determination of the thresholds d50% and d90%. Estimates of the necessary size (longitudinal diameter and short diameter), area and volume to achieve a d50%, and d90% detection rate of 18F-choline PET/CT or 68Ga-PSMA PET/CT are shown in Table 3.
(A–D) Empiric detection rates from 18F-choline PET/CT and the 68Ga-PSMA PET/CT groups for conditions 1 and 2 in dependence on diameter of TD (divided into intervals of 1 mm in length). Dark green line (condition 1) and light green line (condition 2) represent the fitted logistic growth curve model and red line the estimated diameters d90% and d50%, respectively. Longitudinal diameters (A and C) above 10 mm and short diameters (B and D) above 8 mm are shown as 1 group.
Empiric detection rate for condition 1 (A and B) and 2 (C and D) in dependence on diameter of TD. Fitted logistic growth curve models for 18F-choline PET/CT and 68Ga-PSMA PET/CT and P value for difference between the 2 curves are shown.
Estimated Diameters Associated with Detection Rates of d90% and d50%, Respectively, for Both Conditions, Stratified for 18F-Choline PET/CT and 68Ga-PSMA PET/CT
d50% and d90% values tend to be smaller with 68Ga-PSMA PET/CT than with 18F-choline PET/CT. Overall, the difference in detection rates was significant under condition 2. For 18F-choline PET/CT, the results for condition 2 suggest that a longitudinal diameter of 11.2 or 5.5 mm is sufficient to reach a detection probability of 90% or 50%, respectively (corresponding values for the short diameter are 7.4 and 3.3 mm). When 68Ga-PSMA PET/CT is used, the results for condition 2 suggest that a longitudinal diameter of 6.3 or 3.7 mm is sufficient to reach a detection probability of 90% or 50%, respectively. The corresponding values for the short diameter are 4.9 and 2.3 mm.
Table 3 also suggests that we can reach with 68Ga-PSMA PET/CT a detection probability of ≥90% for about half of all LNMs considered and a detection probability of ≥50% for about three quarters. The respective figures for 18F-choline PET/CT (19% and 57%, respectively) indicate a considerably worse performance.
DISCUSSION
Regarding only size of LN at CT or MRI, a large number of existing LNMs will be missed because of the presence of micrometastases (15). Consequently, the pooled sensitivity for CT and MRI was reported to be 42% and 39%, respectively, whereas the pooled specificity was 82% for both CT and MRI (5). PET/CT with 18F-choline analogs such as 18F-fluoroethylcholine or 11C-choline provided respectable results for LN staging of PCa; estimates for sensitivity (38%–98%) and specificity (40%–100%), however, turned out to be fairly variable (7). The superiority of 68Ga-PSMA PET/CT to PET/CT with 18F-choline analogs for the detection of PCa lesions (e.g., LNM) with respect to overall accuracy has already been demonstrated by several studies (12,13,16). However, histopathologic data about the minimum tumor load (e.g., TDs in LNM) triggering a positive PET finding are scarce for both tracers (12,13,16).
By applying semiautomatic LN segmentation software, Giesel et al. assessed the LN volume and size of 68Ga-PSMA PET–positive LNMs in 49 LNMs from 21 men with PCa relapse (17). They reported a mean short and longitudinal axis of 68Ga-PSMA PET–positive LNMs of 5.8 and 9.5 mm. In line with our data, the detection limit was below the CT/MRI criteria of 8–10 mm (15). However, a histologic verification of 68Ga-PSMA PET–positive LNs and a consideration of the TD alone were not reported (17).
Measurements of TD in 34 LNMs (median, 13.6 mm) detected and 19 LNMs (median, 4.3 mm) undetected by PSMA PET/CT from 12 men after lymphadenectomy at RP by Budäus et al. (18) deviate markedly from our data and the data provided by Giesel et al. (17). In that report, the approach of anatomic correlation of PET and histopathology was not, however, reported in detail (18).
Unique features of our investigation are the approach of identification of LNM with a direct link from PET/CT to histopathology (conditions 1 and 2) and the considering of TDs in LNM for estimating the detection rates. A necessary short diameter of TD in LNM of 3.9 mm to reach a d90% with 68Ga-PSMA PET/CT (condition 1, Table 3) impressively underlines the fact that lesions of small size (i.e., well below the scanner spatial resolution of about 6–7 mm) can nevertheless be detected with high sensitivity if target expression is high. To corroborate our results from condition 1 analysis (1 LNM per subregion), we repeated this calculation with an increased sample size, also including subregions with 2 verified LNMs (condition 2). This yielded comparable results.
The performance of both tracers in detecting PCa lesions is correlated to the expression and metabolism of choline transports or the expression of PSMA in the membrane of PCa cells (19–21), for example. Choline transporters (e.g., CHT1s and CTL1) (19,20) are essentially involved when choline is rapidly cleared from the blood pool and incorporated into PCa cells (20); differences of their expression in relapsed and primary PCa are well described (22,23). For PSMA, it is known that about 8% of the primary high-risk PCa patients will not present with a tracer enhancement in a 68Ga-PSMA PET/CT (24,25). With respect to the PSMA study population investigated in 2017 (11), we highlighted the correlation of SUVmax in LNM, the intensity of anti-PSMA immunohistochemistry, and the diameter of TD in the LNM (11): all the LNMs on hematoxylin and eosin staining were clearly positive for anti-PSMA, meaning that no LNM had been missed because of a lack of PSMA expression. For LNMs with a direct link from PET/CT to histopathology, a correlation of SUVmax with the area-weighted PSMA score was not significantly different (r = 0.1536 P = 0.3136; 95% confidence interval, −01.55 to 0.435). In contrast, a correlation of SUVmax with the maximum diameter of the TDs in the LNM was highly significant (r = 0.6720 P < 0.0001; 95% confidence interval, 0.464–0.809) (11).
As we investigated in our work PET-positive patients (PET-positive LNMs) exclusively, however, the heterogeneity mentioned above might be limited. Furthermore, differences of choline transporter and PSMA expression within 1 patient (e.g., between LNMs from 1 primary tumor) are not clearly investigated and probably more unlikely than between individuals. In summary, we suppose that the variation of choline transporter expression or PSMA expression might not be a significant bias in this analysis.
Our results regarding the histopathologic minimum tumor load in LNMs for detection of metastases through 18F-choline or 68Ga-PSMA PET/CT are remarkable by themselves, regardless of the comparison of both tracers. Whether a lymphadenectomy after 68Ga-PSMA PET/CT or 18F-choline PET/CT changes the fate of a patient with nodal metastases remains to be seen. Several more clinical parameters such as timing of surgery, tumor load, and the extent on lymphadenectomy will determine whether a patient will benefit from surgery or not.
Although we could demonstrate that when using PSMA PET/CT, in particular, detection rates can be remarkably high for diameters below the scanner resolution, we are still faced with the fact that small LNMs will be overlooked. After the attempt of removing all potential pelvic LNMs, a target lymphadenectomy in PET-positive regions only will not be universal as PET-negative subregions might harbor small LNMs remaining undetected. A systematic bilateral lymphadenectomy should therefore still be performed. Our data indicate that if available, a 68Ga-PSMA PET/CT should be preferred to 18F-choline PET/CT to reveal LNMs. There are probably additional circumstances (tumor activity in LNM, PSMA-negativity) determining whether an LNM will be detected or not. Of course, it is therefore necessary to improve the detection rate of currently used tracers. Whether an improved detection rate, and thereby an assumed complete surgical resection, or target radiotherapy will lead to a significantly improved outcome in those metastasized patients is yet to be seen.
This study had limitations. Actually, patients with eligible subregion for condition 1, 2 were included in this study, regardless of the character of the clinical data. We therefore had to register the clinical data (such as PSA level and number of LNMs) as they were present at the time of PET imaging with respect to surgery. The aim of Table 1, however, is to highlight that the clinical data from both groups were predominantly balanced.
The number of available LNMs in both cohorts was limited. This is a consequence of the need to identify LNMs for which a direct link to the PET/CT results could be established. Using only subregions with a single LNM is a clear and well-defined approach to establish this link. In the 68Ga-PSMA PET/CT group in particular, however, only a few PET/CT-negative LNMs of this type could be found (reflecting already the superiority of PSMA in detecting small LNMs). We hence enlarged both groups by applying condition 2. This increase in sample size allowed us to demonstrate a significant overall difference in detection rates between 18F-choline and 68Ga-PSMA PET/CT.
LNM samples of the 18F-choline and the 68Ga-PSMA PET/CT analyses originated from different patient populations. This limitation would be circumvented by performing a combined 18F-choline and 68Ga-PSMA PET/CT in a prospective study cohort. However, given the expected limited additional clinical value of an 18F-choline PET/CT scan (when a 68Ga-PSMA PET/CT is available) and the considerably higher radiation, financial, and logistic burden of a dual-tracer approach, we chose to analyze our well-defined earlier patients cohorts.
Because only patients with known BR relapse and the suspicion of LNM on 18F-choline or 68Ga-PSMA PET/CT have been included in this study, our patient cohort is biased toward patients with tumor manifestations that show sufficient 18F-choline uptake and PSMA expression. This, however, is a general prerequisite for 18F-choline and 68Ga-PSMA PET/CT that needs to be kept in mind.
CONCLUSION
In line with earlier studies demonstrating a higher diagnostic accuracy of 68Ga-PSMA PET/CT than PET/CT with 18F-choline analogs for identification of LNMs, the present study suggests the detection limit of TDs is lower for 68Ga-PSMA PET/CT. Whether this results in an improved outcome of patients undergoing 68Ga-PSMA PET/CT–guided lymphadenectomy needs to be investigated further.
DISCLOSURE
No potential conflict of interest relevant to this article was reported.
Footnotes
Published online Jan. 25, 2019.
- © 2019 by the Society of Nuclear Medicine and Molecular Imaging.
REFERENCES
Jump to section
Related Articles
Cited By...
- Pattern of Failure in Patients with Biochemical Recurrence After PSMA Radioguided Surgery
- Head-to-Head Comparison of 68Ga-Prostate-Specific Membrane Antigen PET/CT and Ferumoxtran-10-Enhanced MRI for the Diagnosis of Lymph Node Metastases in Prostate Cancer Patients
- Reply: The Importance of an Adequate Surgical Template During Salvage Lymph Node Dissection for Node-Recurrent Prostate Cancer
- 68Ga-PSMA PET/CT for Primary Lymph Node and Distant Metastasis NM Staging of High-Risk Prostate Cancer
- Multimodality Imaging of Prostate Cancer