Abstract
Repeatable quantification is essential when using 18F-fluoromethylcholine PET/CT to monitor treatment response in prostate cancer. It has been shown that SUV normalized to the area under the blood activity concentration curve (SUVAUC) provides a better correlation with full kinetic analysis than does standard SUV. However, the precision of SUVAUC is not known yet. The purpose of this study was to assess the repeatability of various semiquantitative 18F-fluoromethylcholine parameters in prostate cancer. Methods: Twelve patients (mean age ± SD, 64 ± 8 y) with metastasized prostate cancer underwent two sets of 18F-fluoromethylcholine PET/CT scans, on consecutive days. Each set consisted of a 30-min dynamic PET/CT scan of the chest after intravenous administration of 200 MBq of 18F-fluoromethylcholine, followed by a whole-body PET/CT scan at 40 min. The dynamic scan was used to derive the area under the blood activity concentration curve. Lesion uptake was derived from the whole-body scan using various types of volumes of interest: maximum, peak, and mean. Each of these parameters was normalized to injected activity per body weight, area under the blood activity concentration curve, and blood concentration itself at 40 min, resulting in several types of SUVs: SUV, SUVAUC, and SUVTBR. The test–retest repeatability of these metrics, as well as metabolic tumor volume (MTV) and total uptake of choline in the lesion, were studied. The level of agreement between test–retest data and reliability was assessed using Bland–Altman plots, repeatability coefficients, and intraclass correlation coefficients (ICCs). Results: A total of 67 choline-avid metastases were identified: 44 bone lesions and 23 lymph node lesions. In the case of SUVmax, the repeatability coefficients for SUV, SUVAUC, and SUVTBR were 26% (ICC, 0.95), 31% (ICC, 0.95), and 46% (ICC, 0.89), respectively. Similar values were obtained for SUVpeak and SUVmean. The repeatability of SUVAUC was comparable to that of SUVmax, SUVpeak, and SUVmean. Tissue type and tumor localization did not affect repeatability. An MTV of less than 4.2 cm3 had larger variability than larger volumes (repeatability coefficient, 45% vs. 29%; P = 0.048). The repeatability coefficient did not significantly differ between lesions with SUVpeak above or below the median value of 8.3 (19% vs. 28%; P = 0.264). Conclusion: The repeatability of SUVAUC was comparable to that of standard SUV. The repeatability coefficients of various semiquantitative 18F-fluoromethylcholine parameters (SUV, MTV, and total uptake in the lesion) were approximately 35%. Larger differences are likely to represent treatment effects.
- repeatability
- positron emission tomography (PET)
- prostate cancer
- choline
- standardized uptake value (SUV)
Prostate cancer is the second most common cancer in men worldwide and was the third most diagnosed malignancy in Europe in 2012, with 92,000 deaths (1,2). This androgen-dependent neoplasm is characterized by a good initial response to antihormonal therapy and an unpredictable latent castration-resistant status (3). At the beginning of this decennium, molecular profiling studies improved our knowledge about the heterogeneous biologic behavior of prostate cancer. It was found that even in the presence of a castrate range of serum testosterone (<1.7 nmol/L), a proportion of tumors in castration-resistant prostate cancer patients remains dependent on androgen-receptor signaling for growth (4). Five potential mechanisms of development of castration-resistant prostate cancer were described, based on ligand and androgen-receptor dependence (5).
Today, several therapeutic options against castration-resistant prostate cancer prevail, including cytotoxic (docetaxel, cabazitaxel), antihormonal (abiraterone, enzalutamide), immunotherapeutic (sipuleucel-T), and bone-targeting (223Ra-dichloride) agents (6–13). However, despite this variety of new agents with demonstrated improvement in life expectancy, the proper sequencing (e.g., modality and timing) in individual patients with metastatic prostate cancer is unclear (14). Because therapeutic options vary greatly with the stage and grade of the disease, the specific pattern of metastatic spread (i.e., hematogenous or lymphatic), and the dominant phenotype, accurate diagnostic instruments for response evaluation are essential (15,16).
Noninvasive hybrid PET/CT is a valuable diagnostic tool by virtue of its combining metabolic and anatomic information in vivo (17). Encouraging results have been reported on the usefulness of radiolabeled-choline PET/CT in prostate cancer (18,19). Apart from its main recognized application in restaging disease in cases of biochemical relapse (20,21), 18F-fluoromethylcholine might also qualify as a biomarker of response to therapy. Because, conceptually, choline uptake represents viable tumor cells, tracer uptake changes over time might serve as an improved readout of treatment efficacy.
In vitro experiments have shown promising results for the use of radiolabeled choline to monitor antiandrogen treatment and chemotherapy (22,23). Recently, simplified quantitative methods for measuring 18F-fluoromethylcholine uptake have been developed and validated (24); SUV normalized to the area under the blood activity concentration curve (SUVAUC) correlates better with full kinetic analysis than does standard SUV (24). However, the precision of SUVAUC, and also that of SUV itself, is as yet unknown.
The purpose of this study was to prospectively assess the repeatability of semiquantitative 18F-fluoromethylcholine PET/CT parameters in prostate cancer, also including metabolic tumor volume (MTV) and total uptake of choline in the lesion. Such knowledge is essential for proper interpretation of changes in the 18F-fluoromethylcholine signal over time, thus improving personalized therapy strategies for prostate cancer patients.
MATERIALS AND METHODS
Patients
Twelve patients with histologically proven prostate cancer (4 of whom had castration-resistant prostate cancer) and lymphatic or hematogenous metastases were included prospectively. The inclusion criteria required at least 2 metastases (diameter ≥ 1.5 cm) detected by conventional imaging performed no more than 3 mo before PET/CT and the ability to remain supine for 60 min. The exclusion criteria were claustrophobia and coexistence of multiple malignancies. The study was approved by the Medical Ethics Review Committee of VU University Medical Center. Before inclusion, each patient signed a written informed consent form after receiving a verbal and written explanation of the study.
Personal and demographic data were collected (age, body weight, Gleason score, prostate-specific antigen level (ng/mL) at the time of PET/CT, and information on previous therapy), as well as the characteristics of the metastatic lesions (location [intrathoracic, intraabdominal, or pelvic], number, and type [bone or lymph node]). Values are presented as mean ± SD.
Data Acquisition
All patients underwent 18F-fluoromethylcholine PET/CT on 2 consecutive days. The tracer was synthesized according to a previously published method (24). The minimal interval between the last treatment and the first PET/CT scan was 19 d. Patient preparation was similar to that required for 18F-FDG (25). The patients were scanned using a Gemini TF-64 PET/CT scanner (Philips).
Each patient underwent a low-dose CT scan (30 mAs, 120 kV) followed by a 30-min dynamic PET scan of the chest (24), centered over a large blood-pool structure (e.g., the ascending aorta), to obtain an image-derived input function. At the start of the dynamic 18F-fluoromethylcholine scan, a bolus injection of 205 ± 9 MBq (day 1) and 206 ± 7 MBq (day 2) of 18F-fluoromethylcholine was administered intravenously using an automated injector (Medrad), which was flushed with 40 mL of saline (5 mL at 0.8 mL·s−1 followed by 35 mL at 2 mL·s−1). Dynamic PET data were normalized and corrected for decay, scatter, random coincidences, and photon attenuation and were reconstructed into 25 frames (1 × 10, 8 × 5, 5 × 20, 5 × 60, 3 × 150, and 3 × 300 s) with a matrix size of 144 × 144 × 45 and voxels of 4 × 4 × 4 mm3, using a 3-dimensional row action maximum likelihood algorithm.
After a standard (5-min) micturition break to warrant a proper visual assessment of the pelvic region, a whole-body 18F-fluoromethylcholine scan (mid-thigh to skull base) was performed 40 min after injection. After this PET acquisition (10 bed positions at 2 min each), a second low-dose CT scan (50 mAs, 120 kV) was acquired for anatomic correlation and attenuation correction. Whole-body data were corrected for dead time, decay, scatter, and randoms and were reconstructed into 34 frames (1 × 10, 8 × 5, 4 × 10, 3 × 20, 5 × 30, 5 × 60, 4 × 150, and 4 × 300 s) with a matrix size of 144 × 144 voxels (4 × 4 × 4 mm3), using iterative reconstruction (binary large-object ordered-subsets time-of-fight). The transaxial spatial resolution was about 5 mm full width at half maximum in the center of the field of view, similar to that of the dynamic scan.
Data Analysis
The reconstructed images were transferred to off-line workstations for further analysis. Data were analyzed on a volume-of-interest (VOI) basis (26). The dynamic scan was used to derive the area under the blood activity concentration curve, by defining a cylindric VOI 1.5 cm in diameter extending over 5 consecutive axial planes within the lumen of the ascending aorta. Next, lesion uptake (defined as tracer accumulation exceeding local background activity and incompatible with physiologic tracer biodistribution) was derived from the whole-body scan.
The term metabolic tumor volume was used to indicate volumes that were derived directly from the PET studies, quantified as the VOI size (26). VOIs were defined by a semiautomatic delineation tool, applying a background-adapted 50% of maximum isocontour (i.e., contour value = 50% of maximum + background). SUVmax, SUVpeak, and SUVmean were calculated for each VOI. In addition, each of these parameters was normalized to injected activity per body weight (SUV), area under the blood activity concentration curve (SUVAUC) as derived from the dynamic scan, and blood concentration at 40 min after injection (SUV tumor-to-background ratio [SUVTBR]) as derived from the whole-body scan. For the parameters associated with SUVmean, total choline uptake in the lesion (defined as SUVmean × MTV) was also calculated, resulting in total uptake, area under the total uptake curve, and total uptake tumor-to-background ratio. The test–retest repeatability of all metrics was calculated using both standard repeatability coefficients (1.96 × SD of difference between test and retest) and relative repeatability coefficients (percentage test–retest difference).
Statistical Analysis
Data were analyzed using SPSS, version 15.0 (IBM). Test–retest repeatability was quantified using intraclass correlation coefficients (ICCs; based on absolute agreement) and repeatability coefficients and was displayed graphically using Bland–Altman or box plots. ICCs were calculated for each SUV measure as the ratio of between-lesion variability and total variability (both between-lesion and within-lesion). Variance components for patient, lesion within patient, and repeated measurements within lesion were estimated using mixed models. Total variability was calculated as the sum of these 3 variance components. Between-lesion variability was calculated as the sum of the variance components for patient and for lesion within patient. Confidence intervals for ICCs were determined using the delta method. Reliability coefficients were based on the relative difference:
When repeatability coefficients were calculated, the method of Bland and Altman (27) was used to take into account correlation of measurements between different lesions within the same patient. Differences in the repeatability of the different SUV measures (SUVmax, SUVpeak, and SUVmean) and methods of normalization (SUV, SUVAUC, and SUVTBR) were assessed by comparing the variances of the relative test–retest differences, using the Pitman–Morgan test (28) for correlated variances. Differences in the repeatability of SUV measures related to location, lesion type, and lesion size were assessed by comparing the variances of the relative test–retest differences between lesion subgroups, using the Levene test (29). A Bonferroni correction was used to control overall type I error to 5% within each set of comparisons.
RESULTS
Patients
The 12 patients had a mean age of 64 ± 8 y, a mean body weight of 88 ± 9 kg, a Gleason score of 7 (n = 3) or higher (n = 9), and a median prostate-specific antigen level of 46 ng/mL (range, 2–226 ng/mL) at the time of PET/CT (Table 1). All patients had been treated previously by antihormone therapy, with the addition of chemotherapy alone in 3, chemotherapy plus external-beam radiotherapy in 1, external-beam radiotherapy alone in 2, external-beam radiotherapy plus prostatectomy in 2, external-beam radiotherapy plus immunotherapy in 1, and prostatectomy plus lymph node dissection in 1.
Patient Characteristics
At PET, 67 metastases were identified (median, 6 per patient; range, 3–8), 44 of which were to bone and 23 to lymph node metastases. Twelve metastases were above the diaphragm; the other 55 were intraabdominal or pelvic. The malignant nature of the metastases was confirmed radiologically on the basis of progression of preexistent lesions or new metastatic sites. Median size and interquartile range were 4.9 cm3 and 7.6 cm3, respectively, for VOI and 8.3 and 5.2, respectively, for SUVpeak. Repeatability data were analyzed using a volume threshold of 4.2 cm3 based on a repeatability study of MTV with 18F-FDG and 18F-fluorothymidine in lung cancer (26). In that study, changes in 18F-FDG uptake of more than 37% for lesions larger than 4.2 cm3 were found to represent a biologic effect. This volume threshold corresponds by approximation to a diameter of 2 cm (for spheric metastatic lesions), which equals about 4 times the spatial resolution of PET, below which it is not possible to determine quantification, VOI definition, or detectability without interference by partial-volume effects.
Repeatability of Semiquantitative Parameters
The repeatability of each semiquantitative parameter (i.e., SUV, MTV, and total lesion uptake) was studied as a function of uptake (median SUVpeak of 8.3), MTV (larger or smaller than 4.2 cm3), metastatic tissue type (bone or lymph node), and location (intrathoracic vs. abdominal or pelvic).
Repeatability of SUV
In the case of SUVmax, the repeatability coefficients (for relative differences) were 26% for SUV, 31% for SUVAUC, and 46% for SUVTBR. Similar values were observed for SUVpeak and SUVmean. All ICCs were 0.95 or 0.96 for SUV and SUVAUC and 0.89 for SUVTBR (Table 2).
Test–Retest Differences, ICCs, and Repeatability Coefficients for Different Types of SUVs and Their Normalizations
Nine pairwise comparisons for different methods of SUV normalization were performed to estimate variance in relative test–retest differences. After correction for multiple comparisons (Bonferroni-corrected significant difference, P < 0.0056 at the 5% level and P < 0.0011 at the 1% level), only the SUVTBR parameters were found to have consistently larger variances (P < 0.001) (Supplemental Table 1). The relative percentage differences between test–retest data and the means for SUVmax, SUVpeak, and SUVmean and their normalizations are presented in Figure 1. The repeatability of SUVAUC for SUVmax, SUVpeak, and SUVmean was comparable to that of the corresponding SUV measures (Supplemental Table 2).
Relative differences between test–retest data and mean values for different types of SUVs (SUVmax, SUVmean, and SUVpeak) and their normalizations (SUV, SUVAUC, and SUVTBR). Bonferroni-corrected significance levels are P < 0.0056 (for significance level of 5%, denoted by *) and P < 0.0011 (for significance level of 1%, denoted by **). ● = outliers (>2 SDs); TRT = test–retest.
Repeatability coefficients did not differ between lesions with an SUVpeak above or below the median value of 8.3 (19% vs. 28%; P = 0.264) (Fig. 2 and Supplemental Fig. 1). These values were comparable to those of SUVpeak/AUC (23% [SUVpeak > 8.3] vs. 31% [SUVpeak < 8.3]; P = 0.136). Moreover, the repeatability of SUVpeak and SUVmean was independent of MTV (Fig. 3), as well as of tissue type and tumor location (Supplemental Fig. 2; Supplemental Table 3).
Bland–Altman plots for relative differences between test–retest data and mean values for SUVpeak.
Bland–Altman plots for relative test–retest repeatability of SUVpeak (A) and SUVmean (B), as function of MTV.
Repeatability of MTV
For MTV, the test–retest difference was 0.03 ± 1.63 and the relative test–retest difference 36% (Fig. 4). The repeatability coefficients for MTV were independent of SUVpeak (34.25% [SUVpeak > 8.3] vs. 36.43% [SUVpeak < 8.3]; P = 0.933]. An MTV of less than 4.2 cm3 had larger variability than larger volumes (repeatability coefficient, 45% vs. 29%; P = 0.048) (Supplemental Fig. 1). The repeatability coefficients for MTV did not differ between bone/lymph nodes (34% vs. 36.4%; P = 0.684) and location (36.7% [intrathoracic] vs. 34.4% [intraabdominal/pelvic]; P = 0.820).
Relative differences between test–retest data for MTV as function of MTV.
Repeatability of Total Lesion Uptake
The repeatability coefficients for total lesion uptake and area under the total lesion uptake curve were comparable (33% vs. 31%; P = 0.954), whereas the total lesion uptake tumor-to-background ratio showed a larger variance of 51% (P < 0.001) (Supplemental Fig. 3). No significant difference was found between lesions with a total uptake below or above the median value of 30.9 (repeatability coefficient, 40.9% [<30.9] vs. 23.1% [>30.9]; P = 0.093).
The repeatability coefficients for total lesion uptake were independent of uptake (31% [SUVpeak < 8.3] vs. 34% [SUVpeak > 8.3]; P = 0.139) and MTV (42.1% [<4.2 cm3] vs. 25.3% [>4.2 cm3]; P = 0.037). The repeatability coefficients for total lesion uptake were also independent of tissue type (30.8% [bone] vs. 35.7% [lymph nodes]; P = 0.241) and location (35.5% [intrathoracic] vs. 32.4% [intraabdominal/pelvic]; P = 0.778).
An overview of all repeatability coefficients for the semiquantitative parameters (SUV, MTV, and total lesion uptake) as a function of uptake, MTV, and metastatic tissue type and location is presented in Supplemental Tables 4 and 5. Individual-patient repeatability coefficients for all semiquantitative parameters are provided in Supplemental Figures 4–6. The repeatability coefficients were comparable across most subjects. Patient 1 had the poorest repeatability coefficients regarding SUV (∼35% for all 9 combinations of SUV types and normalizations). Patients 2, 4, and 7 had the poorest repeatability coefficients regarding MTV (∼51% in all 3 patients) and total lesion uptake (55%, 51%, and 67%, respectively).
DISCUSSION
In a previous study, we investigated 18F-fluoromethylcholine kinetics in metastatic prostate cancer (24) and demonstrated that SUV cannot be used to estimate uptake. SUVAUC based on 2 consecutive PET scans was proposed as a clinically feasible alternative. In the present study, we prospectively assessed the repeatability of quantitative 18F-fluoromethylcholine PET/CT parameters in patients with prostate cancer and found that the repeatability of SUVAUC is comparable to that of standard SUV and that differences in uptake of 30% or more are likely to represent treatment effects.
Test–retest repeatability is essential for clinical implementation of any parameter of response assessment. Because of the heterogeneous biologic behavior of prostate cancer (4), and in light of the rapid evolution of treatment modalities (14), biologic markers are needed that adequately monitor response to therapy. The standard treatment-response criteria, RECIST, do not apply to metastatic prostate cancer, further complicating the issue of evaluating response (30). A subsequently proposed system for measuring functional response with 18F-FDG PET/CT, PERCIST, might also apply to radiolabeled choline (31). Nevertheless, before 18F-fluoromethylcholine PET/CT can be implemented as a biomarker for response evaluation in prostate cancer, the repeatability of the tracer should be known (32).
To the best of our knowledge, the repeatability of 18F-fluoromethylcholine measurements in metastatic prostate cancer has not been assessed previously. Pegard et al. (33) addressed the reproducibility of observer interpretation (i.e., visual evaluation and classification of focally increased uptake as being malignant or benign) of 18F-fluoromethylcholine PET/CT examinations in patients with biochemically recurrent prostate cancer. The authors found good concordance when evaluating bone metastases and abdominal or pelvic lymphatic recurrences in previously treated patients. A limited usefulness was found at the prostate level in untreated patients.
Observed repeatabilities in our study were within the range seen for other commonly used radiotracers, such as 18F-FDG and 18F-fluorothymidine. In analyzing changes in 18F-FDG uptake, the generally accepted PERCIST response classification assumes a true biologic change when there is a change in SUVpeak exceeding 30% in combination with a 0.8-unit change in absolute SUVpeak (31). In a metaanalysis on the repeatability of 18F-FDG uptake measurements in tumors, de Langen et al. (34) identified 8 eligible studies. SUVmean had better repeatability than SUVmax. A minimal relative change of 20% in combination with a 1.2-unit change in SUVmean was presumed to represent a biologic change. Comparable results were reported by Rockall et al. (35) in a study on the repeatability of quantitative 18F-FDG PET/CT in recurrent ovarian carcinoma. The repeatability coefficients suggested that a decrease from baseline of up to 20% for tumor SUV and up to 15% for tumor size could be used to determine early tumor response. In a study addressing the repeatability and reproducibility of 18F-FDG and 18F-fluorothymidine PET measurements of tumor volume, Hatt et al. (36) found comparable percentage differences for the two tracer datasets. Differences larger than 30% were considered indicative of treatment response.
Frings et al. (26) analyzed the repeatability of MTV with 18F-FDG and 18F-fluorothymidine in lung cancer. Repeatability was better for larger tumors. For 18F-FDG, changes of more than 37% in lesions larger than 4.2 cm3 represented a biologic effect. We obtained comparable results when using this MTV threshold to analyze 18F-fluoromethylcholine test–retest data. A larger variability was found in small MTVs (<4.2 cm3), suggesting that a similar lower threshold for MTV should be used in treatment-response studies. In the case of total lesion choline uptake, repeatability coefficients were not significantly different for lesions smaller or larger than 4.2 cm3 (after correcting for multiple comparisons; Bonferroni-corrected significance level, P < 0.0056). However, the uncorrected P value equaled 0.037, suggesting a trend toward poorer repeatability for smaller lesions.
Two recently published papers explored the prognostic value of metabolic parameters when 18F-fluoromethylcholine PET/CT is used in biochemically recurrent prostate cancer (37) or castration-resistant prostate cancer (38). In a multivariate analysis, Colombié et al. (37) identified age less than 70 y, SUVmean greater than or equal to 3, and standardized metabolic activity greater than or equal to 23 as independent prognostic factors for disease-free survival. In a prospective study using 18F-fluoromethylcholine PET/CT, Kwee et al. (38) found that whole-body tumor indices based on the quantification of net metabolically active tumor volume and total lesion activity were predictive of overall survival. In our study, we used comparable metrics, with an emphasis on the repeatability of 18F-fluoromethylcholine as a potential biomarker for response evaluation in prostate cancer.
Repeatability coefficients for individual patients were comparable across most subjects. However, patient 1 had the poorest repeatability coefficients regarding SUV, because of a small lesion (1.5-cm short-axis diameter) at the right ventral edge of a vertebral body (Th10), close to the liver. We hypothesize that the relatively large test–retest difference (∼35%) of this lesion was caused by incorrect scatter correction due to high image-derived blood activity concentrations near physiologically 18F-fluoromethylcholine–avid structures. Such an effect might lead to large quantification errors with image-derived input function obtained from blood VOI in these areas (24). Three patients (2, 4, and 7) had larger repeatability coefficients for MTV and total lesion uptake than the other subjects. This difference is likely explained by difficulty in lesion segmentation, such that errors in MTV are also propagated into poorer repeatability for total lesion uptake. Besides, all these patients presented with small metastatic lesions (∼1.5 cm) with slightly increased 18F-fluoromethylcholine uptake.
With respect to the proposed 30% cutoff for SUV measures, the test–retest differences pooled over all lesions or lesions with an SUV of more than 8.3 were approximately 20% (Fig. 2). However, for lesions with an SUV of less than 8.3, the repeatability coefficients were about 30% (Supplemental Table 4). Since, in clinical practice, most patients will have a combination of metastatic lesions with SUVs both above and below 8.3, we decided to adopt the more conservative value of 30%.
A possible limitation of our study is the limited number of subjects. The minimum required sample size was calculated to be 12 patients (with minimally 2 measurements per patient). This sample size yields 80% power for testing the hypothesis ICC ≤ 0.6 against the 1-sided alternative (ICC > 0.6) at a significance level of 5% when the true ICC is equal to 0.9 (39). Moreover, for this sample size of 12 patients, the confidence intervals for the limits of agreement ranged to approximately 1.1 times the SD of the difference scores, at either side of the estimated limit of agreement (27). Thus, the present study provides a reasonable estimate of the expected repeatability coefficients.
CONCLUSION
In patients with metastatic prostate cancer, the repeatability of SUVAUC was comparable to that of standard SUV, indicating that differences in 18F-fluoromethylcholine uptake of 30% or more are likely to represent treatment effects. The repeatability of MTV and total lesion uptake was about 35%. The observed repeatabilities were on the same order of magnitude as those seen for other commonly used radiotracers, such as 18F-FDG and 18F-fluorothymidine.
DISCLOSURE
The costs of publication of this article were defrayed in part by the payment of page charges. Therefore, and solely to indicate this fact, this article is hereby marked “advertisement” in accordance with 18 USC section 1734. No potential conflict of interest relevant to this article was reported.
Acknowledgments
We thank our colleagues at the Department of Radiology and Nuclear Medicine for assistance with tracer production and data acquisition. We are also grateful to the urologists at VU University Medical Center (André N. Vis, MD, PhD) and Amstelland Hospital Amstelveen (Joop W. Noordzij, MD, and Sven Nadorp, MD), as well as the oncologists at VU University Medical Center (Catharina Menke-van der Houven van Oordt, MD, PhD, Joyce van Dodewaard-de Jong, MD, and Jens Voortman, MD, PhD), for assistance with patient selection and inclusion.
Footnotes
Published online Dec. 23, 2015.
- © 2016 by the Society of Nuclear Medicine and Molecular Imaging, Inc.
REFERENCES
- Received for publication October 2, 2015.
- Accepted for publication December 1, 2015.