Dual-Phase PET-CT to Differentiate [18F]Fluoromethylcholine Uptake in Reactive and Malignant Lymph Nodes in Patients with Prostate Cancer

Purpose To investigate whether time-trends of enhanced [18F]Fluoromethylcholine ([18F]FCH) in lymph nodes (LN) of prostate cancer (PCa) patients can help to discriminate reactive from malignant ones, and whether single time point standardized uptake value (SUV) measurements also suffice. Procedures 25 PCa patients with inguinal (presumed benign) and enlarged pelvic LN (presumed malignant) showing enhanced [18F]FCH uptake at dual-phase PET-CT were analyzed. Associations between LN status (benign versus malignant) and SUVmax and SUVmeanA50, determined at 2 min (early) and 30 min (late) post injection, were assessed. We considered two time-trends of [18F]FCH uptake: type A (SUV early > SUV late) and type B (SUV late ≥ SUV early). Histopathology and/or follow-up were used to confirm the assumption that LN with type A pattern are benign, and LN with type B pattern malignant. Results Analysis of 54 nodes showed that LN status, time-trends, and ‘late’ (30 min p.i.) SUVmax and SUVmeanA50 parameters were strongly associated (P<0.0001). SUVmax relative difference was the best LN status predictor. All but one inguinal LN showed a decreasing [18F]FCH uptake over time (pattern A), while 95% of the pelvic nodes presented a stable or increasing uptake (pattern B) type. Conclusions Time-trends of enhanced [18F]FCH uptake can help to characterize lymph nodes in prostate cancer patients. Single time-point SUV measurements, 30 min p.i., may be a reasonable alternative for predicting benign versus malignant status of lymph nodes, but this remains to be validated in non-enlarged pelvic lymph nodes.


Introduction
Choline-based PET has proven value in several neoplasms, and especially in prostate cancer [1,2]. [ 18 F]FCH PET-CT seems to become a useful imaging tool to solve the clinical problem of rising serum prostate-specific antigen (PSA) after initially treated prostate cancer [3][4][5], with PET-CT sensitivity being proportionally related to the PSA level [6,7].
Choline is the precursor of phosphatidylcholine (PC), an essential phospholipid of cell membranes. The phosphorylation process is catalyzed by choline kinase (CK) [2]. In prostate cancer, enhanced choline uptake is explained by both increased mitogenic activity as well as up regulation of CK [8]. However, the signal is not tumor specific [9,10]. Since biopsy of presumed regional lymph node metastases is often not trivial due to their localization in prostate cancer, false positive readings may induce serious problems in the clinical context.
Both [ 18 F] and [ 11 C]-labeled choline derivates have been developed and studied as possible metabolic imaging tools in the detection of primary PCa, regional LN and distant metastases [11,12]. The main advantages of [ 18 F] over [ 11 C]-labeled tracers are the longer half-life (110 min versus 20 min) and a better spatial resolution due to the shorter positron range of 18 F. However, [ 18 F]choline is excreted in urine [13] and this can compromise the interpretation of the pelvic area. Since [ 18 F]FCH is rapidly cleared from the blood pool, acquisition protocols have been designed with imaging of the pelvis prior to bladder filling (within minutes after injection), followed by a whole body scan after e.g. 30 min [1,[14][15][16] -the dual-phase protocol. Patterns of tracer uptake as a function of time have shown to be helpful in discriminating intraprostatic tracer uptake [8,15] as well as sites of suspected haematogeneous metastases [17].
It was suggested that increasing or stable [ 18 F]FCH uptake over time (3-7 min versus 30-60 min after injection) was compatible with malignancy, while a decreasing tracer uptake is associated with benign status [8,15,17]. For lymph node assessment this hypothesis has not been validated. Beheshti et al. [8] described an interesting early [ 18 F]FCH wash-out pattern of false-positive LN in the preoperative setting of 130 patients with high and intermediate risk for disseminated prostate cancer. This phenomenon was deemed highly important for the differentiation of malignant versus benign nodes. However, the study could not support this finding with a statistically relevant number of lesions.
In the present study we investigated whether time-trends of enhanced [ 18 F]FCH uptake in lymph nodes can help to discriminate between benign and malignant sites, and we explored whether single time point SUV measurements may also suffice. Patient characteristics [age, Gleason score (Gl), serum PSA at diagnosis and at the time of performing the PET-CT] were gathered, including the date and the type of previous therapy [e.g. radical prostatectomy (RP), external-beam radiation therapy (EBRT), brachytherapy (BT), pelvic lymph node dissection (PLND), anti-androgen therapy (ADT) or the combination of these].

Patients
Inclusion criteria were: dual-phase [ 18 F]FCH PET-CT performed in patients with histopathologically proven PCa; enhanced [ 18 F]FCH uptake in any inguinal nodes and in pelvic LN with a short axis diameter $8 mm, visible at early and/or late PET scan. Patients with multiple malignancies (e.g. one or more other types of carcinoma apart from the PCa) were excluded.
PSA relapse, suspected to be associated with residual or recurrent disease, was defined as a serum concentration level above 0.2 ng/ml, after RP and after the combination of RP with other types of therapy. A rising PSA level .2 ng/ml above the nadir value in patients treated by means of EBRT was considered suspect for persistent or recurrent disease [18]. Increased serum PSA was confirmed by two consecutive exams for all patients. The maximal time interval between performing the PET-CT scan and the last PSA determination was 14 days.

Lymph Node Classification
We classified lymph nodes as benign or malignant using the following approach: inguinal lymph nodes with enhanced [ 18 F]FCH uptake were considered benign since the prostatic lymph node drainage pattern does not include inguinal nodes [20][21][22]. Pelvic nodes with a short axis diameter $8mm were classified as malignant [23]. Additional confirmation was obtained using histopathology (whenever feasible), and with clinical follow-up of 6-12 months in all patients. Follow-up consisted of PSA measurements over time (as above mentioned) and/or evaluation of other imaging (i.e., contrast-enhanced abdominopelvic CT, pelvic MRI). We applied the following radiological criteria to classify change [24]: an increase by 30% versus initial size as progression, a decrease by 30% as regression, and intermediate values or no change in size as stable. Therefore, malignancy was defined as a positive histopathological result; radiologically confirmed progression in size of the pertaining lymph nodes; decrease or normalization of serum PSA and radiological response after therapy, providing these were the only abnormal findings on the initial PET-CT scan; decrease or normalization of serum PSA with nodal regression after RT, with the RT field including the site of the suspected LN.

Synthesis of [ 18 F]FCH
[ 18 F]FCH was synthesized according to the methods proposed by DeGrado et al. [1,25] and Iwata et al. [26], with minor modifications and by use of automated modules built in-house [27]. In short, cyclotron produced [ 18 F]fluoride was reacted with dibromomethane, the formed [ 18 F]fluorobromomethane was purified and used in the alkylation of 2-(dimethylamino)ethanol to obtain 2.660.9 GBq [ 18 F]FCH after semi-preparative high performance liquid chromatography (HPLC) purification, reformulation in aqueous 0.9% NaCl and sterile filtration.
The radiochemical purity was .99.5% and no chemical impurities were detected as assessed by analytical radio/ultraviolet-HPLC (UV-HPLC). Residual concentrations of dibromomethane (0-8 ppm) and 2-(dimethylamino)ethanol (95640 ppm) were determined by flame ionization detector -gas chromatography (FID-GC). Absence of bacterial endotoxins in the product was confirmed by an Endosafe portable test system (PTS) reader (Charles River) and all samples were tested for sterility.

PET-CT Imaging
All studies were performed on a Gemini TOF-64 PET-CT scanner (Philips Medical Systems, Best, The Netherlands) with an axial field per view of 18 cm. Low-dose CT (LD-CT) was collected using a beam current of 30 to 50 mAs at 120 keV. CT was reconstructed using an image matrix size of 5126512 resulting in voxel sizes of 1.1761.17 mm and a slice thickness of 5 mm. For PET, data were reconstructed by means of a raw action ordered subset expectation maximization algorithm using default reconstruction parameters. Time of flight (TOF) information was used during reconstruction. Reconstructed images had an image matrix size of 1446144, a voxel size of 464 mm and a slice thickness of 5 mm.
All patients underwent the standard [ 18 F]FCH image acquisition protocol at our institution: following the LD-CT, 'early' PET image acquisition started 2 min after intravenous injection of 4 MBq/kg [ 18 F]FCH, using a 35 cm scan trajectory over the pelvic region (2 min/bed position); patients were asked to void 20 min post injection (p.i.), and at 30 min p.i. a 'late' whole body PET sequence was started, from mid-thigh to the skull vertex, again using 2 min acquisitions/bed position. Patient preparation was similar to that required for FDG PET [28].

PET-CT Data Analysis
PET-CT images were evaluated by an experienced nuclear medicine specialist who first identified all lymph nodes with enhanced FCH uptake versus their direct background (on early and/or late scan time points) within the field of view of the early scan trajectory. Lymph node diameters were measured using the CT component of the PET-CT scanner, and standard CT and MRI, where available.
PET and LD-CT images were converted to ECAT 7 format and regions of interest (ROIs) were semi-automatically drawn around every pelvic LN that met the inclusion criteria mentioned above, using in-house developed software, as previously described [29,30]. For lesion delineation, we used the adaptive threshold of 50% of maximum voxel value within tumor, the 3D volume of interest A50 (VOIA50). This method is similar to the fixed threshold method, except that it adapts the threshold relative to the local average background, thereby correcting for the contrast between tumour and local background. For example, the A50 contour value corresponds to a value at 50% of the sum of the maximum voxel value and the local background value. The latter value is derived from 'background' voxels that are identified as those voxels located on a single voxel thick shell at 2.5 cm from the edge of a 70% of maximum pixel value isocontour, excluding all voxels with an SUV larger than 2.5.
SUV's were normalized for body weight. For data-analysis, we used early (2 min p.i.) and late (

Statistical Analysis
The two sample Mann-Whitney tests were performed to determine a shift in the median values for benign and malignant tumors. Linear mixed-effects models were constructed to determine the relation between SUV max and SUV mean . Included in this model were tumor status (benign versus malignant) and post injection scan time (2 min versus 30 min) as fixed effects, as well as random slopes per patient and per lesion (nested within patient). The residuals were assumed to be exponential related to SUV mean , and pairwise interactions between SUV mean and both tumor status and scan time were also tested. An identical linear regression model ignoring repeated measures and heteroscedasticity was used to provide an R 2 . Receiver Operating Characteristic (ROC) curves were constructed to determine the thresholds maximizing specificity and sensitivity. The binomial distribution is used to determine 95% confidence intervals of the sensitivity and specificity estimates for the optimal thresholds. To ensure that these results are not influenced by within-patient correlations in uptake, a Monty Carlo process is performed where ROC curves are produced for 500 datasets randomly generated so that each patient has only one lesion. The range of the sensitivities and specificities of the optimal thresholds from these ROC curves were visually compared with the sensitivities and specificities calculated using all observations.

Results
We identified 25 eligible patients who had 54 lymph nodes with enhanced [18F]FCH uptake that met our inclusion criteria. In 13/  Table 1.
All LN showed enhanced FCH uptake at early (2 min p.i.) as well as late (30 min p.i.) time points ( Table 2). The long axes of the inguinal and pelvic lymph nodes were similar (1162, and 1263 mm, respectively); short axis diameters were slightly smaller in the former group (861 vs. 1062 mm, respectively; P,0.001).
We found highly significant associations between the LN status (inguinal/benign vs. enlarged pelvic/malignant) and the SUV max and SUV meanA50 30 min p.i., and their absolute and relative differences (P,0.0001). The correlation between the mean and max SUV metrics was near-perfect ( Fig. 1; the linear regression model resulted in an almost identical relation and an adjusted R 2 of 0.99). For further analyses we focused on the SUV max . The number of LN with a [ 18 F]FCH uptake pattern type A (SUV early . SUV late ) or B (SUV early # SUV late ), in the benign and malignant group for the above mentioned thresholds are included in Table 3. Based on the SUV max relative difference, all but one LN in the benign group showed a type A, decreasing uptake over time, while in the malignant group 95% (35/37) of the nodes presented a type B pattern uptake. From the 54 LN, only 3 nodes were found to have a stable uptake over time: one from the inguinal and two from the pelvic group, respectively. ROC analyses of uptake trends over time and of SUV's at either time-point (Fig. 2) showed that the SUV max relative difference was the best predictor of the lymph node status, followed by the SUV max late and the SUV max absolute difference [see also Table 3 for the areas under the ROC curve (AUC)]. The threshold of SUV max relative difference that maximized both sensitivity and specificity was 20.19, for a sensitivity of 95% (95%CI 82-99) and a specificity of 94% (95%CI 71-100), versus a SUV max late threshold of 2.3, for a sensitivity of 92% (95%CI 78-98) and a specificity of 88% (95%CI 64-99).
In the 17 patients with 34 enlarged [ 18 F]FCH positive pelvic nodes, histopathological confirmation was obtained in 4 (11 LN); all had type B time trends at [ 18 F]FCH PET-CT. During followup, 3 patients (6 LN) had radiological nodal progression; from these five LN had a type B pattern and one LN proved to be false negative: type A trend and radiological progression. In 3 other patients, with only pathologically enlarged pelvic lymph nodes (6 LN) to explain an elevated PSA and with a type B pattern, PSA normalized upon therapy, accompanied by shrinkage of these nodes. PSA decrease and disappearance of a solitary (type B) pelvic node was observed in another patient who was treated with pelvic radiotherapy that focused on a suspected recurrence in prostate and this lymph node.
Six other patients (with 9 type B and 1 type A pelvic lymph nodes) were treated with systemic therapy, because of local recurrence and/or skeletal metastases. Finally, the patient with both 2 inguinal and 3 pelvic LN (type A in the inguinal and type B in the pelvic nodes) was treated with RT (prostate) and ADT because of confirmed bone oligometastases at presentation. In these 7 patients, PSA decreased over time and radiological regression of all 13 enlarged pelvic nodes together with stable dimensions of the inguinal nodes were recorded, as well. However, this did not unequivocally confirm the LN status due to the use of systemic therapy for coexistent bone metastases.
Of the inguinal nodes (14 type A and 1 type B with stable uptake over time) none showed signs of malignancy during follow- up; of these, in one patient a lymph node biopsy was tumor negative. One patient (2 LN) opted for watchful waiting policy (otherwise negative PET), and his PSA was stable at 0.4 ng/ml for the follow-up of 12 months. In another patient (type A LN) a local recurrence was treated with HIFU with good clinical response (PSA) and no signs of progression.

Discussion
In our referral-based spectrum of patients with enhanced [ 18 F]FCH uptake in pelvic and inguinal lymph nodes, decreasing  We classified disease-status as malignant for enlarged pelvic nodes, and as benign for inguinal nodes of any size. Our criterion of benignity was based on the typical prostate drainage pattern which does not include inguinal nodes, as described by Inoue et al. [20]. They identified by using fluorescence navigation 3 lymphatic drainage pathways, comprising the obturator, the external and internal iliac nodes. Similar drainage patterns were found by Tokuda et al. [21] in 125 patients with LN metastases. Weckermann et al. [22] performed both sentinel lymph node dissection and radical prostatectomy in 1055 patients with PCa. Despite a high percentage (.50%) of positive nodes identified outside the standard lymphadenectomy borders, none of them were found in the inguinal region. In our study we also never encountered occurrences of malignancy in inguinal nodes (histological analysis, clinical radiological follow-up).
We considered pelvic nodes with a short axis diameter equal or exceeding 8 mm as being malignant. This threshold was chosen based on the study of Jager et al. [23] who reported a 98% specificity for MRI using this dimension. In their meta-analysis, Hövels et al. [31] found that false positivity of CT/MRI (similar performance for either technique) at thresholds of 8-10 mm is only 7%. In an attempt to reduce the remaining uncertainty, a standard of reference method was used. This approach, as extensively described in the ''Materials and Methods'' section,  consisted of the combination of histopathological examination (whenever available) and the results obtained by clinical or radiological follow-up. This is a commonly used procedure [32][33][34][35][36][37] to account for the limitations of retrospective studies. In difficult cases, biopsy of the proper radioactive choline avid lymph nodes was improved and verified by using a dedicated gammaprobe [38]. Confirmation seemed feasible in 65% of these pelvic LN (24/37). In 7 patients treated with ADT and/or chemotherapy for coexisting bone metastases, decreases of nodal diameter could not be interpreted since such changes are not necessarily compatible with a 'malignant tissue' response to treatment.
Note that in our present context 'sensitivity' and 'specificity' should not be confused with 'the accuracy of [ 18 F]FCH PET-CT to diagnose metastatic lymph nodes in prostate cancer'. The results pertain to the ability of tracer uptake time-trends to classify lymph nodes with enhanced [ 18 F]FCH uptake.
The relevance of uptake time-trends to characterize [ 18 F]FCH foci has been demonstrated in malignant bone metastases, in recurrent PCa, and in malignant zones of the prostate in preoperative setting [8,15,17]. Our findings corroborate and extend those of Beheshti et al. [8] who reported on 18 malignant lymph nodes showing stable or increasing uptake over time. The imaging protocol consisted of a dynamic PET/CT scan of the pelvic region for 8 min, starting 1 min p.i., followed by whole body (WB) images 10 min after [ 18 F]FCH injection and optional supplementary delayed WB acquisitions, 90-120 min p.i., when abnormalities were detected. However, since that study comprised only 4 [ 18 F]FCH positive reactive lymph nodes (with decreasing uptake over time) they urged for validation of these patterns in a larger study.
In our study, all but one inguinal nodes showed decreasing [ 18 F]FCH uptake over time (Fig. 3), versus 95% (35/37) of the pelvic category demonstrating stable or increasing uptake (Fig. 4). Kwee et al. [15] suggested as a possible explanation for the tracer decrease over time in benign zones the dephosphorylation of [ 18 F]FCH by prostatic acid phosphatase while a trapping mechanism of the tracer in the malignant cells was responsible for the increased uptake in PCa. This can only be validated with full kinetic modeling.
Our results are at variance with those of Cimitan et al. [15] who reported no significant late/early [ 18 F]FCH uptake ratios in proven local recurrent prostatic disease or abdominopelvic LN, when performing a dual phase [ 18 F]FCH PET-CT in 43 patients with PSA relapse. However, their dual-phase PET/CT protocol included WB scans with variable early and late acquisitions, 5 to 15 min p.i. and 65 to 200 min p.i., respectively. The discrepancy may be explained by the rapid clearance of the [ 18 F]FCH after administration [14,16], which implies that timing of the early acquisitions is crucial and should be done using a strict imaging protocol.
The prevalence of patients with inguinal lymph nodes showing enhanced [ 18 F]FCH uptake was 12% (8/66 patients). Our finding that reactive nodes remained detectable over time (i.e., for 30 min. after injection) seems being inconsistent with the observation of Price et al. [16] who found persistent uptake at the late, 45 min p.i., images in only 1 of 4 patients with initially enhanced inguinal node uptake. We content that this variance results from improved signal to noise ratios with the current TOF scanner generations.
Our data suggest that the type B pattern is a strong indication for malignancy, the PPV being 97% (35/36; see Table 3, for the best LN status predictor: SUVmax Relative Difference). The fact that the PPV is not 100% implies in clinical practice that, e.g. in case of multiple potentially malignant [ 18 F]FCH positive LN, the ones with type B patterns should be the primary biopsy candidates.
Our results also suggest that a single point measurement in the context of a whole body scan, starting in caudocranial direction, 30 min after injection, is a reasonable alternative for relative change to differentiate reactive from malignant LN in patients with PCa. Obviously, omitting the early scan would simplify the scan procedure. However, compared to measuring relative changes, the use of absolute SUV's is more demanding at the level of standardization. Evidently, the present results pertain to lymph nodes with enhanced [ 18 F]FCH uptake and $8 mm short axis. Ascertaining the single time-point approach requires validation versus histopathology.

Conclusion
Time-trends of enhanced [ 18 F]FCH uptake in lymph nodes of prostate cancer patients seem to help discriminate benign from malignant localizations. Single time-point SUV measurements, 30 min p.i., may be a reasonable alternative for predicting the nodal status, but this remains to be validated in non-enlarged pelvic lymph nodes.

Author Contributions
Conceived and designed the experiments: DEOL RJAvM WRG AJMvdE OSH. Performed the experiments: DEOL ADV OSH. Analyzed the data: DEOL ADV RJAvM WRG JE RB OSH. Contributed reagents/ materials/analysis tools: ADV AJMvdE JE RB. Wrote the paper: DEOL ADV RJAvM WRG AJMvdE JE RB OSH. Authored the statistical parts in the drafts: ADV. Provided raw data for analysis: RJAvM WRG AJMvdE. Provided expert support concerning choline: JE. Provided technical support for generating data: RB. Project coordinator: OSH.