Metabolic Volumetric Parameters in 11C-Choline PET/MR Are Superior PET Imaging Biomarkers for Primary High-Risk Prostate Cancer

Purpose Positron emission tomography/magnetic resonance imaging (PET/MRI) can facilitate the use of noninvasive imaging biomarkers in clinical prostate cancer staging. Although multiparametric MRI is a widely used technique, the clinical value of simultaneous PET imaging remains unclear. This study aimed at investigating this issue. Methods Between January 2015 and December 2016, 31 high-risk prostate cancer patients underwent 11C-choline PET/MRI for staging purposes. Clinical characteristics and imaging parameters, including the standardized uptake value (SUV) and metabolic volumetric parameters from PET imaging; apparent diffusion coefficient (ADC) values from diffusion-weighted imaging; and volume transfer rate constant (Ktrans), reflux rate constant (Kep), and initial area under curve (iAUC) in 60 seconds from dynamic contrast-enhanced (DCE) MRI were analyzed. Results 11C-Choline PET imaging parameters were significantly correlated with prostate-specific antigen (PSA) levels, and metabolic volumetric parameters, including metabolic tumor volume (MTV) and uptake volume product (UVP), showed significant correlations with other MRI parameters. In our cohort analysis, the PET/MRI parameters UVP/minimal ADC value (ADCmin) and kurtosis of Kep (Kepkur)/ADCmin were significant predictors for progression-free survival (PFS) (HR = 1.01, 95% CI: 1.00–1.02, p=0.031 and HR = 1.09, 95% CI: 1.02–1.16, p=0.009, respectively) in multivariate Cox regression analysis. High UVP/ADCmin and Kepkur/ADCmin values were significantly associated with shorter PFS. Conclusions Metabolic volumetric parameters such as MTV and UVP can be routinely used as PET imaging biomarkers to add prognostic value and show better correlations in combination with MR imaging parameters in high-risk prostate cancer patients undergoing 11C-choline PET/MRI.


Introduction
Prostate cancer (PCa) is the leading cause of cancer-related death among men in Taiwan, with its incidence increasing rapidly from 26.22 per 100,000 males in 2002 to 47.86 per 100,000 males in 2012 [1]. e conventional imaging modalities used to assess PCa include computed tomography (CT), magnetic resonance imaging (MRI), transrectal ultrasound (TRUS), and bone scintigraphy. However, simultaneous positron emission tomography (PET)/MRI has great potential to enhance clinical practice in these cases by combining functional, molecular, and anatomic information [2].
Although 11 C-choline has been approved for patients with suspected PCa recurrence by the U.S. FDA, the combination of 11 C-choline and MRI may also be effective for staging, especially in patients with a high Gleason score, advanced clinical stage, and elevated prostate-specific antigen (PSA) levels [3,4]. Biomarkers derived from medical images offer several advantages, including ready availability, noninvasiveness, and serial patient monitoring [5]. e apparent diffusion coefficient (ADC) in diffusion-weighted MRI and the standardized uptake value (SUV) in PET have been used together as an imaging biomarker in several previous studies [6][7][8]. However, studies on the use of 11 C-choline PET metabolic volumetric parameters in combination with the SUV and ADCs and dynamic contrast-enhanced (DCE) parameters from multiparametric prostate MRI, which may theoretically maximize the value of simultaneous PET/MRI assessments, in cases of high-risk PCa are still lacking. Herein, we sought to identify clinically significant integrated 11 C-choline PET/MRI parameters for high-risk primary PCa by assessing the correlations of these parameters with clinical characteristics and progression-free survival (PFS).

Study Patients.
is was a retrospective analysis of a prospective study. e study was approved by the hospital's institutional review board (approval numbers: 102-3271A and 201701793B0; ClinicalTrials.gov identifier: NCT02852122), and informed consent was obtained from all patients. Between January 2015 and December 2016, 54 consecutive patients with a clinical indication for PCa staging were scheduled to undergo an 11 C-choline PET/MRI examination, and a total of 31 patients were eventually assessed in this retrospective analysis. e study scheme is shown in Figure 1. All 31 patients had pathologically proven highrisk prostate cancer according to the D'Amico Risk Classification [9].
We recorded primary treatments (i.e., radical prostatectomy, hormone therapy, and radiation therapy plus hormone therapy) for prostate cancer in all 31 patients, and their follow-up status after treatments. e follow-up period was until June 2018. Disease progression after primary treatments was considered if the patients showed local or nodal relapse, new metastasis, or biochemical failure (a PSA value ≥0.2 ng/ml with a secondary confirmatory PSA value >0.2 ng/ml after radical prostatectomy; PSA value >2 ng/ml above nadir after radiation therapy without or with hormone therapy; or a continuous increase in PSA values after hormone therapy) [10][11][12]. e follow-up durations from diagnosis to disease progression or death/last visit of all patients were recorded.

11 C-Choline PET/MRI Protocol.
After fasting for at least 6 h, patients received a single intravenous bolus of 10-20 mCi (370-740 MBq) 11 C-choline; the mean dose was 16.5 ± 3.6 mCi. Approximately 5 min after 11 C-choline injection and bladder evacuation, whole-body PET/MRI scanning was performed using an integrated PET/MRI system (Biograph mMR; Siemens Healthcare, Erlangen, Germany). PET scans were performed from the mid-thigh to the head in five bed positions (acquisition time, 3 min per position) with the patient in a supine arms-down position. Simultaneous MRI was performed with a transverse T2-weighted half-Fourier single-shot TSE (turbo spin-echo) sequence (1,000 ms repetition time (TR)/84 ms echo time (TE), 6 mm slice thickness, 320 × 256 matrix, and 380 × 309 mm 2 field of view (FOV)) and a coronal T1-weighted TSE sequence (500 ms TR/9.5 ms TE, 5 mm slice thickness, 1.5 mm intersection gap, 384 × 276 matrix, and 450 × 310 mm 2 FOV), while acquiring PET data in each bed position. e simultaneous whole-body PET/MRI acquisition was followed by pelvic PET/MRI scans ( Figure 2) involving a pelvic PET scan in one bed position (emission period, 15 min). e MRI pulse sequences included a sagittal T2weighted TSE sequence (4,000 ms TR/91 ms TE, 4 mm slice thickness, 320 × 224 matrix, and 200 × 200 mm 2 FOV), coronal T2-weighted TSE sequence (4,000 ms TR/80 ms TE, 4 mm slice thickness, 0.4 mm intersection gap, 256 × 179 matrix, and 180 × 177 mm 2 FOV), and transverse T2weighted TSE sequence (3,600 ms TR/80 ms TE, 4 mm slice thickness, 0.4 mm intersection gap, 256 × 179 matrix, and 180 × 177 mm 2 FOV), and acquisitions were performed in the pelvic region. Axial diffusion-weighted imaging (DWI) was performed using a single-shot spin-echo echoplanar imaging technique under free-breathing conditions (5,000 ms TR/65 ms TE, b values of 50 and 1000 s/mm 2 , 4 mm slice thickness, 106 × 106 matrix, 260 × 205 mm 2 FOV, and NEX 6). Axial DCE-MRI was performed using a 3D T1weighted spoiled gradient-echo sequence (3.91 ms TR/1.6 ms TE, 4 mm slice thickness, 128 × 128 matrix, 256 × 200 mm 2 FOV, and 13°FA). A total of 72 volumes were acquired with a temporal resolution of 4.16 seconds and acquisition time of 5 minutes. After four acquisitions of dynamic baseline scanning, a standard dose (0.1 mmol/kg body weight) of gadopentetate dimeglumine (Gd-DTPA; magnevist; Bayer-Schering, Burgess Hill, UK) was administered by a power injector through a cannula placed in the antecubital vein at a rate of 3 mL/s and immediately followed by a saline flush.
Attenuation correction of the PET data was performed using a four-tissue (air, lung, fat, and soft tissue) segmented attenuation map acquired with a two-point Dixon MRI sequence. Images were reconstructed using a high-definition PET (HD-PET) iterative algorithm (three iterations; 21 subsets) with a 5.4 mm post-reconstruction Gaussian filter and an image matrix of 344 × 344.

PET Imaging Parameter Analysis.
e PMOD 3.3 software package (PMOD Technologies Ltd., Zurich, Switzerland) was used for tumor segmentation. A volume of interest was manually drawn around the PCa lesion using an SUV cutoff of 2.5 in accordance with previous studies [13,14]. PET imaging features were calculated in an SUV analysis. e maximum SUV (SUV max ), mean SUV (SUV mean ), metabolic tumor volume (MTV), and uptake volume product (UVP) were derived according to the following equations: SUV � (tissue radioactivity/tissue weight (g))/(total radioactivity (Bq)/body weight (g)) and UVP � SUV mean × MTV. Computations for imaging features were performed using the CGITA (Chang Gung Image Texture Analysis) toolbox implemented using MATLAB 2012a (MathWorks, Inc., Natick, MA, USA) [15].

MR Imaging Parameter Analysis.
Postprocessing for DWI and DCE-MRI was performed using the software integral to the MRI unit (Siemens Syngo Via and Tissue 4D; software version, VA20B). e ADC was calculated from the diffusion-weighted images. e Toft model [16] was used for pharmacokinetic analysis to derive the following parameters from the DCE-MRI data: the volume transfer rate constant (Ktrans), reflux rate constant (Kep), and initial area under curve (iAUC) in 60 seconds. e ADCand the DCE-MRI-derived parameter maps were reconstructed on a pixel-by-pixel basis. On axial DCE images, an uroradiologist with over 20 years of experience manually drew the largest region of interest (ROI) within each primary tumor on each image ( Figure 3). ROIs on axial ADC images were obtained using a similar approach with the aid of diffusion-weighted images. e histogram data of the derived parameters from the ROIs-including maximal value and kurtosis of Ktrans (Ktrans max and Ktrans kur ), Kep (Kep max and Kep kur ), and iAUC (iAUC max and iAUC kur )-and the minimal value, mean value, and kurtosis of ADC (ADC min , ADC mean , and ADC kur ) were exported for statistical analysis using a homemade software written in MATLAB (R2015b; MathWorks, Inc., Natick, MA, USA).

Fifty-four patients agreed to participate in this study
Did not meet inclusion criteria (n = 9) Informed consent withdrawn: 4 patients Unable to undergo PET/MR due to claustrophobia: 2 patients No histological proof of prostate cancer: 1 patient Androgen deprivation therapy before PET/MR: 2 patients  Contrast Media & Molecular Imaging 2.5. Statistical Analysis. Descriptive statistics were calculated to summarize the data, using the median (range) for continuous variables and count (percentage of total) for categorical variables. Intergroup comparisons of continuous variables were based on the Mann-Whitney U-test. Correlations among study variables were investigated using the Spearman's correlation method. PFS was defined as the interval between diagnosis and disease progression. Univariate and stepwise multivariate Cox regression analyses based on the forward Wald method, with thresholds of 0.05 and 0.1, respectively, for entering and removing variables, were performed to identify predictors significantly associated with PFS. Two-sided p values less than 0.05 were considered statistically significant.

Patient Clinical Characteristics.
e clinical characteristics and PET/MR imaging parameters are listed in Table 1. More than half of the patients were aged ≥70 years. PSA values of 20-50 ng/ml and Gleason scores of 7 were the most common. Clinical T4 tumors were present in more than half of the patients. e majority of patients had clinical stage IV disease, and hypertension was the most common comorbidity. Table 2 shows the correlations between PET and MRI parameters. Among the PET imaging parameters, MTV and UVP more frequently showed significant correlations with DCE and ADC parameters. MTV was significantly correlated with Ktrans max , Kep max , Kep kur , iAUC max , and iAUC kur (σ � 0.39-0.51, all p < 0.05), whereas UVP was significantly correlated with Ktrans max , iAUC max , and iAUC kur (σ � 0.37-0.51, all p < 0.05). SUV max showed moderate and significant correlation with iAUC max . Other SUV and DCE parameters showed no significant correlations. Both MTV and UVP were negatively correlated with ADC min and positively correlated with ADC kur . Other SUV and ADC parameters showed no significant correlations. ere were no significant correlations between DCE and ADC parameters, except between Ktrans max and ADC kur (σ � 0.37, p < 0.05).

Correlations among PET/MR Parameters, PSA Levels, and Gleason Scores.
PSA levels were positively correlated with PET imaging parameters, including SUV max , SUV mean , UVP, and MTV, and negatively correlated with ADC min and ADC mean of MRI. Gleason scores showed positive and significant correlations with both Kep kur and iAUC kur .

PET/MR Imaging Parameters Stratified by Clinical
Characteristics.
e relevant intergroup comparisons of imaging parameters are shown in supplementary Tables S1-S3. All PET parameters but none of the MRI parameters were associated with PSA ≥20 ng/ml (all p < 0.05).
ree MRI parameters (Kep kur , iAUC kur , and ADC mean ) were associated with Gleason scores of 8-10 (all p < 0.05). Multiple PET/MRI parameters were associated with the disease stage T3-4, including MTV and UVP from PET, Ktrans max , and Kep max from DCE-MRI and ADC min and ADC kur from ADC assessments in MRI (all p < 0.05). In contrast, none of the PET/MRI parameters showed significant differences between patients without and with regional lymph node metastasis. Similarly, many PET/MRI parameters (i.e., MTV, UVP, iAUC kur , and ADC min ) showed associations with both distant metastasis and stage IV disease (all p < 0.05). SUV max was also associated with distant metastasis, and ADC mean , with comorbidities (both p < 0.05).

Patients' Disease Progression Predictions by Imaging
Parameters. Of the 31 patients, 4 were not treated at our hospital and 27 had undergone radical prostatectomy (n � 6), radiation therapy (n � 2), hormone therapy (n � 11), or radiation plus hormone therapies (n � 8

Discussion
Unlike biospecimen-based markers, imaging biomarkers undergo technical, biological, and clinical validation along with assessments of cost-effectiveness before they are routinely used in clinical settings. Although PET/MR offers robust data due to the high soft tissue contrast and the addition of multiparametric MRI improves its ability to evaluate prostate cancer in T-staging, the improvements offered by this approach in the detection of nodal disease or bone involvement appear uncertain [17,18]. Metabolic volumetric PET parameters such as MTV and UVP are considered better than SUV because they provide more accurate PCa characterization [13]. Increased MTV and UVP values reflect the volume of viable tumor cells and high tissue metabolism. In the current study, we performed a comprehensive analysis of the correlations among imaging biomarkers as well as between imaging biomarkers and clinical parameters. e findings showed that MTV and UVP are significantly correlated with many DCE-MRI parameters, which provide tumor blood microcirculationrelated information.

Contrast Media & Molecular Imaging
Ktrans has been used to quantitatively assess microvascular permeability; Kep reflects the rate of contrast agent transfer from the extravascular extracellular space back to the blood [16]; and iAUC represents the general tumor blood flow, overall perfusion, and tumor interstitial space index. ese parameters are correlated with angiogenesis [19,20]. e correlations among these parameters are not surprising since primary tumors with higher T stages such as ≥T3 have higher MTV, UVP, Ktrans max , and Kep max values (Supplementary Tables S1 and S2). Furthermore, MTV and UVP values are also significantly correlated with ADC min . ADC values reflect the degree of water restriction, and the increased cellularity of PCa increases water restriction and reduces ADC values. Similarly, patients with distant metastasis also have higher MTV and UVP values but lower ADC min values (Supplementary Table S3). us, higher MTV and UVP values may reflect the tumor characteristics of high-risk PCa, not only the higher metabolic activity of larger viable tumors but also the higher angiogenesis and water restriction noted on MRI, and may be useful in categorizing PCa patients by indicating a high likelihood of a higher primary T stage and distant metastasis.
Several promising radiotracers-some of them targeting choline and prostate-specific membrane antigen (PSMA)are currently being investigated for PET imaging of PCa [21]. An increased cellular membrane synthesis represents a biological substrate for PCa imaging. Specifically, choline enters the cell via the choline transporter, being further phosphorylated to phosphatidylcholine by choline kinase. Both of these molecules are upregulated in tumor cells, ultimately resulting in an enhanced choline uptake [22]. 11 C-Choline PET imaging parameters, including SUV max , SUV mean , MTV, and UVP, are significantly correlated with the PSA level, but none of them are correlated with the Gleason score. is may be because several benign entities, especially benign prostate hyperplasia, will also show increased 11 C-choline uptake [23]. In our data, Kep kur and iAUC kur were correlated with the Gleason score. Kurtosis, as a parameter of first-order histogram analysis, reflects the outliers of a probability distribution in the tail extremity and is considered to be related to tumor heterogeneity [24][25][26]. Prostate cancers typically have high Ktrans, Kep, and iAUC values on DCE-MRI, and both Ktrans and Kep affect iAUC since the iAUC reflects the overall blood volume of PCa in the initial 60 seconds [27]. e positive correlation of Kep kur and iAUC kur on DCE-MRI with the Gleason score implies that PCa with a more heterogeneous distribution of contrast back to the blood and overall blood volume is more likely to have higher Gleason scores. us, 11 C-choline PET/MRI, in combination with DCE-MRI parameters, provides additional information regarding tumor heterogeneity and Gleason scores of prostate cancers.
Consistently with a previous study conducted with 18 Fcholine [28], we failed to identify a significant inverse association between SUV and ADC values. In contrast, our cases showed a significant inverse correlation between ADC min and UVP/MTV. Decreased ADC values indicate increased water restriction as a result of increased cellularity, and increased MTV and UVP values reflect the volume of viable tumor cells and high tissue metabolism. is correlation may be explained by the fact that histologic tumor volume is significantly associated with both ADC min and metabolic volumetric parameters (as shown in a study conducted with 18 F-choline) [13]. Combinations of PET and MRI parameters have been considered as imaging   Contrast Media & Molecular Imaging biomarkers with the prognostic value in many studies. In one study, the MTV/ADC min ratio was found to be an independent predictor of PFS in pancreatic cancer [29].
Another study demonstrated that percentage changes in SUV max /ADC min and tumor lesion glycolysis (TLG)/ADC min can predict treatment response to neoadjuvant chemotherapy (NAC) early in the course of breast cancer treatment [30]. Since multiparametric MRI is a routine clinical tool for prostate cancer diagnosis, the possible prognostic predictive value of PET/MR must be considered. In our study, both UVP/ADC min and MTV/ADC min ratio were significant predictors for disease progression in univariate analysis, and UVP/ADC min was a significant predictor for disease progression in multivariate analysis as Kep kur /ADC min ( Figure S1). Our findings need to be interpreted within the context of some limitations. First, most high-risk PCa patients did not undergo prostatectomy, and the number of patients with surgical specimens was small. Although Gleason scores from TRUS biopsies are acceptably accurate in predicting malignancy, 25%-30% of cases may show discrepancies [31]. Second, the treatment was not standardized in all patients.
ird, the follow-up durations were relatively short (mean follow-up duration, 25.5 months). ese limitations present some uncertainties for the correlation between imaging parameters and biospecimen-based markers and the prognostic significance. However, our study also has several strengths. First, it was the 11 C-choline PET/MR study focusing on the high-risk PCa staging setting. Second, comprehensive imaging analysis, including PET, DWI, and DCE-MRI, was performed. ird, our data suggested that metabolic volumetric PET parameters, including MTV and UVP, are superior to SUV, show significant correlations with ADC and DCE values, and have prognostic value when combined with ADC min . Data Availability e data used to support the findings of this study are included within the article.

Conflicts of Interest
e authors declare that there are no conflicts of interest.

Authors' Contributions
Jing-Ren Tseng and Lan-Yan Yang contributed equally to this work.

Acknowledgments
is study was financially supported by grants CPRPG1G0011, CPRPG1G0012, and CPRP1G0013 from the Chang Gung Memorial Hospital. e authors acknowledge the statistical assistance provided by the Clinical Trial Center, Chang Gung Memorial Hospital, Linkou, Taiwan (funded by the Ministry of Health and Welfare of Taiwan; grant MOHW107-TDU-B-212-123005).

Supplementary Materials
Tables S1, S2, and S3 demonstrate the relationship between imaging parameters and clinical risk features in PET, DCE, and ADC, respectively. Figure S1 shows Kaplan-Meier plots of progression-free survival according to the hybrid imaging parameters (a) UVP/ADC min and (b) Kep kur /ADC min . Comparisons were made with the log-rank test. Youden's index was used to determine the optimal cutoff values based on the area under the receiver operating characteristic curves for the events of interest. (Supplementary Materials)