The Sensitivity of [¹¹C]Choline PET/CT to Localize Prostate Cancer Depends on the Tumor Configuration

Prostate cancer (PCa) is currently the highest prevalent form of cancer in men (192,280 cases, 25% of all incident cases) and constitutes the second most common cause of cancer deaths (9%) in the United States (1, 2). The definite diagnosis depends on the presence of PCa in prostate biopsy cores. Unfortunately the false-negative detection rate for transrectal ultrasound (TRUS)-guided biopsies is exceeding 20%, resulting in repeated biopsies in high-risk patients and in a lasting uncertainty in those patients about the cause of the PSA elevation (3).

Therefore, a diagnostic modality being able to accurately detect and localize PCa in the prostate would be desirable. Differences in the biologic behavior of PCa and other nonmalignant conditions that affect the prostate, such as prostatitis and benign prostate hyperplasia (BPH) are important considerations in the approach to imaging. Imaging modalities like computed tomography, conventional MRI, and TRUS have shown limited accuracy for the diagnosis and staging of primary PCa (4–8). Therefore, a major challenge for imaging undetected local PCa is to increase diagnostic performance.

In recent years, positron emission tomography/computed tomography (PET/CT) has been introduced that combines functional and morphologic data and allows for whole-body imaging. For molecular imaging with PET/CT, a high tumor to nontumor uptake ratio in the prostate is a prerequisite for a PET tracer used for diagnosis of primary PCa. As PET with [¹⁸F]FDG has a limited sensitivity for imaging PCa (9, 10), the choline metabolic pathway has been suggested to be a promising imaging target for PCa (11–15). Although there is good agreement among several centers on the utility of PET/CT with [¹¹C]choline for the detection of recurrent disease after radical prostatectomy in patients with biochemical failure, controversies exist about the accuracy of this technique with respect to the primary detection of PCa. In PET and PET/CT studies that related [¹¹C]choline maximum standardized uptake values (SUV_max) within different prostate regions to the presence of cancer cells in the correspondent sextants either supported (16, 17) or did not support (18, 19) the use of radiolabeled choline to accurately localize primary PCa. Among several factors, tumor configuration has been discussed as a confounding factor impairing the diagnostic accuracy of [¹¹C]choline PET/CT. Therefore, we tested the hypothesis that the diagnostic accuracy of [¹¹C]choline PET/CT to localize PCa depends on the tumor configuration.

Forty-three consecutive patients (median age: 66, range: 50–76 years) with biopsy-proven, untreated PCa were included in this study. Time between biopsy and [¹¹C]choline PET/CT had to be at least 15 days to be included in the study. All patients underwent standardized radical prostatectomy and pelvic lymph node dissection within 31 days (median: 6.0, range: 1–31 days) after [¹¹C]choline PET/CT. In this study we focus on the detection of PCa within the prostate by using histopathology as gold standard. The median time interval between biopsy and PET/CT scan was 34.5 days (range: 15–117 days). The study was approved by the ethics committee. All patients provided informed consent for participation in the study. Patients who had proven concomitant cancer were not included in the study.

[¹¹C]choline was synthesized according to the method of Pascali and colleagues (20) with minor modifications. [¹¹C]choline was produced with radiochemical yields of 80% to 90%, based on [¹¹C]MeI, and radiochemical purity of >99%.

Patients fasted at least 6 hours before [¹¹C]choline PET/CT scanning. Five minutes after injection of 682 ± 75 MBq [¹¹C]choline, patients underwent [¹¹C]choline PET/CT (mid thigh to thorax) on a Sensation 16 Biograph PET/CT scanner (Siemens). The acquisition protocol included sequentially a low-dose CT (26 mA, 120 kV, 0.5 second per rotation, 5-mm slice thickness) for attenuation correction, followed by the PET scan and a diagnostic CT in portal venous phase 80 seconds after i.v. injection of contrast agent (Imeron 300; 240 mA, 120 kV, 0.5 second per rotation, 5-mm slice thickness). All patients received a rectal filling with a negative contrast agent (100–150 mL). All PET scans were acquired in 3-dimensional (3D) mode with an acquisition time of 3 minutes per bed position. Transaxial and axial resolution by a ramp filter are 6.3 and 6.5 mm, respectively, in full-width at half-maximum (FWHM; ref. 21). Emission data were corrected for randoms, dead time, scatter and attenuation and were reconstructed iteratively by an ordered subsets expectation maximization algorithm (4 iterations, 8 subsets) followed by a postreconstruction smoothing Gaussian filter (5 mm FWHM).

Images were analyzed by 2 experienced nuclear medicine physicians and a doctorial student simultaneously, all of them being blinded to the pathology results. For image analysis the transaxial PET slices and the corresponding fused low-dose CT slices were used. Latter were applied to determine the delineation of the organ contour, helping the matching with histology. Corresponding to the histopathologic evaluation, a sextant-based analysis of the images was carried out by using a grid, which was placed on each transaxial slice of the prostate. The first slice evaluated was always the one following the image in which the base of the prostate first appeared. Each prostate slice, except of the first and the last slice evaluated, was divided by the grid into 12 segments—6 central and 6 peripheral (segmental model A; Fig. 1). The first and the last slice of each patient's prostate were divided in 6 segments. The grid was adapted to each slice. To allow an assessment of the amount of tracer uptake we evaluated the [¹¹C]choline uptake by semiquantitative analysis. To minimize influence of the partial volume effect a complementary, second analysis was carried out in which 2 consecutive axial slices were added in one, and the segments of the right and left prostate site in those slices were added together, creating one segment for each side of the prostate in every slice (segmental model B). In each segment of these analyses a region of interest (ROI) was placed and the SUV_max was calculated.

To exclude a potential excretion of the tracer in the bladder as a factor confounding the results, the number of patients exhibiting [¹¹C]choline activity in the bladder was evaluated.

Volumetry of the prostate was carried out on the diagnostic CT by using a software (Volume; Siemens Healthcare) that implements a semiautomatic segmentation algorithm for volume measurements. The user defines an ROI around each axial slice of the prostate and sets a lower and an upper threshold [we chose 5 and 1,000 Hounsfield unit (HU), respectively]. The algorithm creates a volume of interest. All voxels in the volume of interest with HU values between the lower and upper threshold are used for calculating the final volume (22).

The analysis was carried out by a board-certified surgical pathologist with 10 years experience in urologic pathology (mainly prostate). He was blinded to the image results.

After surgical resection, the prostate gland was prepared for histologic evaluation in a way that histopathologic evaluation could be corresponded to the transaxial PET/CT images. The orientation of the prostate was preserved by inking right and left margins with different colors, by anatomic landmarks, and by the different extensions of the prostate slices. After coating with india ink for definition of the resection status (R-status) and for laterality and fixation in 10% buffered formalin, whole-mount axial cross-sections were obtained at 5-mm intervals transversely in a plane perpendicular to the long axis of the gland in craniocaudal direction (base to apex) to match the corresponding axial PET/CT slices. A 4-μm whole-mount section from each slice was then stained with hematoxylin and eosin according to standard procedures. The sections were divided in 12 segments by applying and adjusting the same grid used for the evaluation of the PET/CT images. The presence and location of cancer foci, high-grade prostate intraepithelial neoplasm (HGPIN), prostatitis, and BPH were determined by the pathologist for each sextant. Additionally, the predominant tumor configuration in the histologic specimen was classified in 4 groups: I, unifocal larger than 5 mm (large unifocal); II, multifocal; III, rind-like shaped; and IV, size <5 mm. Tumors were staged according to the classification system of the International Union Against Cancer (UICC) 2004, and graded according to the Gleason score system.

Data analysis included the following:

1. The detection of PCa by SUV_max was investigated. For this purpose the location of the SUV_max of the whole prostate was assessed and compared with histopathology findings. SUV_max was considered as being located in tumor tissue if the tumor site in histopathology was present in the same segment of the prostate within one section. Additionally, we looked for whether the SUV_max was located in the same side (right or left lobe) of the prostate as the tumor in histopathology.

2. The number of cases in which the SUV_max was in a segment adjacent to a segment showing cancer cells in histopathology was determined.

3. Furthermore, we evaluated whether SUV_max was influenced by T-stage, PSA level, and Gleason score.

4. The influence of potential confounding factors such as tumor configuration, T stage, PSA, and Gleason score on tumor prediction was assessed.

5. We evaluated whether the SUV enables to discriminate cancer tissue from BPH, prostatitis, HGPIN, or normal prostate tissue.

Generalized estimation equation (GEE) models (23) were used to assess the explanatory capability of SUV_max for tumor prediction under consideration of tumor configuration, T stage, PSA level, and Gleason score and the ability to discriminate cancer tissue from BPH, prostatitis, and HGPIN in the model. The GEE approach properly reflects the structure of repeated data and takes into account correlation of several segments per individual. Spearman correlation coefficient (ρ) was used to quantify bivariate relationship of quantitative data and to assess statistical significance of monotonous dependencies (test for trend).

To correct for potential increment of false significant results by increased number of hypothesis formally tested, Bonferroni correction of P values was applied in multiple comparisons. All statistical tests were conducted 2-sided and (corrected) P values less than 0.05 were considered to indicate statistical significance.

Histopathology confirmed T₂ (37 of 43; 86%) or T₃ (6 of 43; 14%) PCa in all patients. The tumors were classified according to their configuration in histology as I in 21 patients, as II in 9, as III in 5, and as IV in 8 (Fig. 2). Median PSA value at the time of the PET/CT scan was 6.8 ng/mL (range: 1.0–38.7 ng/mL) and median Gleason score was 6 (range: 5–9; Table 1). Four patients had lymph node metastases, 1 lymph node metastasis each. For more detailed information [TNM stage (tumor node metastasis, system for staging cancer), size of the metastatic lymph nodes, etc.] about these patients see Table 2. [¹¹C]choline PET/CT suggested nodal metastasis in one of these patients (patient no. 4 in Table 2). Overall, [¹¹C]choline PET/CT showed evidence of distant metastases in 2 of 43 patients, both in bone, 1 in the first lumbal vertebral body and 1 in the right acetabulum.

Median volume of the studied prostate glands was 62 mL (range: 30–132 mL). In segmental model A, each inner segment included approximately 2 to 4 voxels, each outer segment approximately 4 to 9 voxels, voxel volume being 0.14 mL.

Overall, in segmental model A, 2,526 segments were analyzed, with a median of 60 segments/patient (range: 30–96 segments/patient). PCa tissue was present in 602 of 2,526 segments (23.8%), BPH in 1,820 of 2,526 (72%), prostatitis in 576 of 2,526 (22.8%), and HGPIN in 149 of 2,526 (5.9%). In 21 of 149 segments (14%) HGPIN was concomitant with PCa, in 34 of 149 (23%) with BPH, in 63 of 149 (42%) with PCa and BPH, in 2 of 149 (1%) with prostatitis, in 9 of 149 (6%) with PCa and prostatitis, in 12 of 149 (8%) with BPH and prostatitis, 2 of 149 (1%) segments solely contained HGPIN, and in 6 of 149 (4%) all the 4 entities were present.

The median SUV_max for all segments was 4.5 (range: 1.4–18.4).

SUV_max was located in the same side of the prostate (right or left) in 41 of 43 (95%) of the patients. The 2 patients in whom SUV_max was not ipsilateral to the tumor were both staged as T₂ and classified as IV (small tumors) and III (rind-like shaped), respectively.

Overall, SUV_max was not significantly associated with PSA (ρ = 0.099; P = 0.526) or Gleason score (test for trend: P = 0.29; Fig. 3A and B). SUV in PCa tissue was significantly higher in T₃-staged tumors (median SUV_max: 6.2, range: 2.5–18.4) compared with T₂-staged tumors (median SUV_max: 4.3, range: 1.4–10.2; P = 0.013).

The area with the highest choline uptake in the prostate did not correspond to a segment involved with PCa in 15 of 43 (35%) of the patients. SUV_max was located in a segment adjacent to a segment involved with PCa in 7 of 15 patients.

Overall, through segmental model B, 214 segments were partied with a median of 4 segments/patient (range: 4–8 segments/patient). Applying segmental model B, the area with the highest choline uptake in the prostate did not correspond to a segment involved with PCa in 9 of 43 (21%) of the patients. All those patients had cancers staged as T₂. SUV_max was not located in PCa in 7 of 8 classified as IV (small tumors), 1 of 5 classified as III (rind-like shaped), and 1 of 9 classified as II (multifocal) in the histologic specimen. SUV_max was located in PCa in all tumors (21 of 21) classified as I (large unifocal) in the histologic specimen.

In 4 of 43 patients, moderate [¹¹C]choline uptake in the bladder was observed. However, SUV was not higher in the segments closer to the bladder. In 3 of these 4 patients, 2 having tumors classified as I and 1 as II, SUV_max was located in the segment involved with PCa. In all 3 cases the segments were not close to the bladder. In 1 patient (tumor classified as IV), SUV_max was not in the segment involved with PCa and also not in a segment close to the bladder.

Tumors classified as IV were significantly less accurately predicted compared with tumors classified as I (P < 0.001; Table 3). T stage, PSA, and Gleason score did not significantly influence tumor prediction (P = 0.24, P = 0.26, P = 0.18, respectively; Table 3). See Table 3 for the results of this analysis about segmental model A.

For assessing whether the different intraprostatic pathologies can be discriminated by the SUV, segments containing only one of the entities were taken into account. Solely PCa was observed in 86 segments, solely BPH in 1,004 segments, and solely prostatitis in 95 segments. Solely HGPIN was observed in 2 segments and was therefore excluded from further analysis. In 454 segments, solely normal prostate tissue was observed. See Tables 4 and 5 for SUV_max of the different entities. After correction for multiple comparisons SUV_max in segments containing solely PCa tissue was significantly higher than that in segments containing normal prostate tissue (P = 0.012) but there was no significant difference when compared with segments containing prostatitis or BPH (P = 0.054 and P = 0.102, respectively). Segments containing BPH had a significant higher SUV_max than segments containing normal prostate tissue (P = 0.042). There was no significant difference between segments containing BPH and those containing prostatitis or between segments containing prostatitis and those containing normal prostate tissue (P > 0.99).

The results of this study indicate that PCa cannot be accurately detected by [¹¹C]choline PET/CT. SUV_max of the prostate was not located in PCa in 21% of patients. The lack of sensitivity of [¹¹C]choline PET/CT to diagnose primary PCa is related to the inability to detect small tumors. Furthermore, a multivariable analysis failed to show that [¹¹C]choline PET/CT can disentangle cancer tissue from benign entities, particularly from BPH, which is due to just minor differences in SUV levels of the entities.

Initial studies utilizing [¹¹C]- or [¹⁸F]-labeled choline derivates and PET reported encouraging results for detection of primary PCa (17, 24). Further, Reske and colleagues, in a study including 26 patients with biopsy-proven PCa, comparing on a segment basis respective transaxial [¹¹C]choline PET/CT and histopathology slices, concluded that major territories with PCa can be imaged precisely, located, and differentiated from benign tissue with [¹¹C]choline PET/CT (16). In contrast, other groups reported on substantially lower sensitivities for [¹¹C]choline PET/CT to diagnose primary PCa [Farsad and colleagues (18), 66%; Martorana and colleagues (25), 66%; Giovacchini and colleagues (26), 72%].

Because of the limited spatial resolution of the scanner and the low number of voxels included in the segments, results are likely to be influenced by the partial volume effect and spillover. There is evidence for such an influence in our study, as—using segmental model A—PCa cells were located in an adjacent segment to the segment exhibiting the highest activity in 7 of 15 patients in whom PCa could not be localized by SUV_max. Therefore, segmental model B, which divided the prostate in substantially larger segments compared with segmental model A, is more appropriate for comparison with histopathology.

Another potential limitation is spillover from activity in the bladder. [¹¹C]choline is mainly not excreted renally. However, moderate activity in the bladder has been observed (27). In our study, as PCa could be localized correctly in 3 of the 4 patients who exhibited some [¹¹C]choline excretion in the bladder and SUV_max was not located in segments closer to the bladder in any of the 4 patients, there was no substantial influence of bladder activity in localizing PCa.

Discrepant results in different studies may be partly due to differences in the patient populations studied. To enable comparability, we discuss the results of segmental model A to outline the differences. BPH was present in 72% of segments in our study, part of them mixed with other entities, 40% of segments contained solely BPH tissue. In the patient population, Reske and colleagues studied a presence of BPH mixed with other entities in 48% of segments is reported, whereas 16% of segments contained solely BPH tissue (16). Farsad and colleagues reported a diffuse and constant presence of BPH in almost all prostatic regions of their studied patients, making them to omit on a distinct analysis about the differentiation of PCa from BPH (18). We omitted an attempt for visual analysis of the images because we observed diffuse intense uptake in the prostate of most patients and lack of a circumscribable focal uptake in many of them. Comparison with histopathology revealed that this was due to the widespread presence of BPH in our patient population. Interestingly the entity being more difficult to be differentiated from PCa in our study was BPH, whereas SUV_max of segments with prostatitis had a tendency to be lower than SUV_max of segments with PCa. Our result that SUV_max of segments with PCa was not significantly different from SUV_max of segments with prostatitis and significantly different when compared with segments with normal prostate tissue could be due to the relatively low number of segments with prostatitis compared with segments with normal prostate tissue. [¹¹C]choline uptake in prostatitis was numerically lower than the uptake of normal prostate and showed the same SD. A higher SUV in prostatitis than in normal tissue could be expected. This was not the case in our study, and apart of the relatively low number of segments with prostatitis, low floridity of the inflammation could be considered. Our results yielding the inability of [¹¹C]choline PET/CT to differentiate PCa from benign prostatic pathologies are in line with the study of Martorana and colleagues (25).

Concerning laterality, [¹¹C]choline PET/CT correctly localized the tumor in 95% of the patients, suggesting that it could be of help in men at risk of having PCa indicated by chronically elevated PSA levels, but experiencing repeated negative biopsies. However, as only men with biopsy-proven PCa have been included in our study, further studies are needed to evaluate this potential indication. Considering 4 of the 43 patients with lymph node metastases, [¹¹C]choline PET/CT suggested nodal metastatic disease in 1. The failure to detect the metastases can be attributed to the small size of the involved lymph nodes in at least 2 of the 3 patients with a false-negative scan.

Similar to previous studies, no significant differences were observed when the SUV was related to PSA or to Gleason score (16, 18, 26, 28), whereas other groups reported of significant correlation between SUV and PSA level (24, 25). One of the potential utilities of [¹¹C]choline PET/CT in primary PCa could be to identify patients with more aggressive lesions to avoid sampling error and appropriately direct therapy for higher grade, riskier tumors. The lack of correlation of uptake with PSA and Gleason score indicates that [¹¹C]choline PET/CT is unlikely to identify these tumors accurately. However, patients included in this study had tumors of relatively lower grade, only 7 of 43 (16%) exhibiting a Gleason score more than 7. Considering that, a definite conclusion on whether [¹¹C]choline PET/CT can provide this information cannot be drawn by our study.

SUV was significantly higher in T₃ than in T₂ tumors in our study. This is in line with Reske and colleagues (16). Notably, patients with T₃ or T₄ tumors represented a higher fraction in the patient population of that study (10 of 26; 38%) compared with the respective fraction in the present study (6 of 43; 14%), giving an additional hint about differences in the patient populations included. The low number of patients with T₃ tumors included, and the fact that the tumors of the patients included were of relatively lower grade, could, in part, explain the detection inefficiency of [¹¹C]choline PET/CT in our study. However, our study indicates, concerning the widespread appearance of BPH in our patient population and exhibiting an uptake that cannot be distinguished from the uptake of PCa, that detection inefficiency of [¹¹C]choline PET/CT is not restricted to tumors of small size but is additionally hampered by the presence of concomitant benign disease, in particular of BPH.

Our study has limitations, which affect all studies comparing corresponding transaxial [¹¹C]choline PET/CT and histopathologic slices. The accuracy of the correlation of findings in respective slices is limited because of image fusion, prostate shrinkage ex vivo due to fixation, tissue distortion generated by the whole prostate step sections (cutting), and a potential sampling error because of the different slice thickness of the 2 methods (5 mm vs. 4 μm).

A further potential limitation is that [¹¹C]choline PET/CT was carried out after prostate biopsy. Although we did not include patients who had undergone a biopsy less than 15 days prior to the PET/CT scan, we cannot exclude that reparative changes after biopsy might have caused false-positive [¹¹C]choline uptake. These may no longer be apparent on histopathologic examination as surgery was done several days later.

Additionally, the precision of the activity measurement in tumors is limited by the partial volume effect and the spillover of activity from neighboring segments. We analyzed and minimized the influence of such factors potentially confounding the results. We excluded spillover from the bladder.

The detection and localization of PCa in the prostate with [¹¹C]choline PET/CT is affected by tumor configuration. Small tumors cannot be visualized. Additionally, in our patient population, PCa tissue could not be distinguished from benign pathologies in the prostate, particularly not from BPH. Therefore, our data do not support the routine use of PET/CT with [¹¹C]choline as a first-line screening procedure for PCa in men at risk. A potential application of [¹¹C]choline PET/CT may be to increase the detection rate of cancer on repeated biopsies in patients who have a persistently high risk of PCa and who have undergone multiple, iterative TRUS-guided biopsies with negative findings.

No potential conflicts of interest were disclosed.

The costs of publication of this article were defrayed in part by the payment of page charges. This article must therefore be hereby marked advertisement in accordance with 18 U.S.C. Section 1734 solely to indicate this fact.