[Skip to Content]
Access to paid content on this site is currently suspended due to excessive activity being detected from your IP address 34.235.143.190. Please contact the publisher to request reinstatement.
[Skip to Content Landing]
Figure 1.
Patient Flow Diagram
Patient Flow Diagram

AR indicates androgen receptor; [18F]-FDG, 2-fluoro-2-D-deoxyglucose F 18; [18F]-FDHT, fluorodihydrotestosterone F 18; Glyc, glycolysis; mCRPC, metastatic castration-resistant prostate cancer; and PET/CT, positron emission tomography/computed tomography. AR0 and Glyc0 indicate negative and AR1 and Glyc1 indicate positive for [18F]-FDHT and [18F]-FDG, respectively.

Figure 2.
Heterogeneity of Lesion and Patient Phenotypes in Metastatic Castration-Resistant Prostate Cancer (mCRPC)
Heterogeneity of Lesion and Patient Phenotypes in Metastatic Castration-Resistant Prostate Cancer (mCRPC)

The distribution of androgen receptor (AR) and glycolysis (Glyc) lesion phenotypes within each of the 133 patients with mCRPC. The left vertical axis and the middle 2 color bar axes represent the fraction of the 3 lesion phenotypes per patient. The right vertical axis and black dots represent the number of lesions per patient. Four patient phenotypes are depicted: (1) concordant (26%): only AR1Glyc1 lesions; (2) AR predominant (25%): AR1Glyc1 and AR1Glyc0 lesions; (3) glycolysis predominant (30%): AR1Glyc1 and AR0Glyc1 lesions; and (4) mixed (19%): mix of all 3 lesion phenotypes. AR0 and Glyc0 indicate negative and AR1 and Glyc1 indicate positive for fluorodihydrotestosterone F 18 and 2-fluoro-2-D-deoxyglucose F 18, respectively.

Figure 3.
Correlations With Overall Survival of 133 Patients
Correlations With Overall Survival of 133 Patients

A, Total metabolizing lesions per patient (median, 12). B, Androgen receptor (AR) and glycolysis (Glyc) AR1Glyc1-only lesions per patient (median number, 7.5). C, Maximum standardized uptake value (SUVmax) of hottest lesion on 2-fluoro-2-D-deoxyglucose F 18 ([18F]-FDG) positron emission tomography/computed tomography (median SUVmax, 7.6). AR1Glyc1 indicates positive for fluorodihydrotestosterone F 18 and [18F]-FDG, respectively.

1.
Siegel  RL, Miller  KD, Jemal  A.  Cancer statistics, 2017.  CA Cancer J Clin. 2017;67(1):7-30.PubMedGoogle ScholarCrossref
2.
Chen  Y, Sawyers  CL, Scher  HI.  Targeting the androgen receptor pathway in prostate cancer.  Curr Opin Pharmacol. 2008;8(4):440-448.PubMedGoogle ScholarCrossref
3.
de Bono  JS, Logothetis  CJ, Molina  A,  et al; COU-AA-301 Investigators.  Abiraterone and increased survival in metastatic prostate cancer.  N Engl J Med. 2011;364(21):1995-2005.PubMedGoogle ScholarCrossref
4.
Fizazi  K, Scher  HI, Molina  A,  et al; COU-AA-301 Investigators.  Abiraterone acetate for treatment of metastatic castration-resistant prostate cancer: final overall survival analysis of the COU-AA-301 randomised, double-blind, placebo-controlled phase 3 study.  Lancet Oncol. 2012;13(10):983-992.PubMedGoogle ScholarCrossref
5.
Scher  HI, Fizazi  K, Saad  F,  et al; AFFIRM Investigators.  Increased survival with enzalutamide in prostate cancer after chemotherapy.  N Engl J Med. 2012;367(13):1187-1197.PubMedGoogle ScholarCrossref
6.
Beer  TM, Armstrong  AJ, Rathkopf  DE,  et al; PREVAIL Investigators.  Enzalutamide in metastatic prostate cancer before chemotherapy.  N Engl J Med. 2014;371(5):424-433.PubMedGoogle ScholarCrossref
7.
Scher  HI, Morris  MJ, Basch  E, Heller  G.  End points and outcomes in castration-resistant prostate cancer: from clinical trials to clinical practice.  J Clin Oncol. 2011;29(27):3695-3704.PubMedGoogle ScholarCrossref
8.
Larson  SM, Morris  M, Gunther  I,  et al.  Tumor localization of 16β- 18F-fluoro-5α-dihydrotestosterone versus 18F-FDG  in patients with progressive, metastatic prostate cancer.  J Nucl Med. 2004;45(3):366-373.PubMedGoogle Scholar
9.
Zanzonico  PB, Finn  R, Pentlow  KS,  et al.  PET-based radiation dosimetry in man of 18F-fluorodihydrotestosterone, a new radiotracer for imaging prostate cancer.  J Nucl Med. 2004;45(11):1966-1971.PubMedGoogle Scholar
10.
Fox  JJ, Autran-Blanc  E, Morris  MJ,  et al.  Practical approach for comparative analysis of multilesion molecular imaging using a semiautomated program for PET/CT.  J Nucl Med. 2011;52(11):1727-1732.PubMedGoogle ScholarCrossref
11.
Gönen  M, Schrag  D, Weiser  MR.  Nodal staging score: a tool to assess adequate staging of node-negative colon cancer.  J Clin Oncol. 2009;27(36):6166-6171.PubMedGoogle ScholarCrossref
12.
Pandit-Taskar  N, O’Donoghue  JA, Durack  JC,  et al.  A phase I/II study for analytic validation of 89Zr-J591 immunoPET as a molecular imaging agent for metastatic prostate cancer.  Clin Cancer Res. 2015;21(23):5277-5285.PubMedGoogle ScholarCrossref
13.
Halabi  S, Lin  CY, Kelly  WK,  et al.  Updated prognostic model for predicting overall survival in first-line chemotherapy for patients with metastatic castration-resistant prostate cancer.  J Clin Oncol. 2014;32(7):671-677.PubMedGoogle ScholarCrossref
14.
Robinson  D, Van Allen  EM, Wu  YM,  et al.  Integrative clinical genomics of advanced prostate cancer.  Cell. 2015;161(5):1215-1228.PubMedGoogle ScholarCrossref
15.
Scher  HI, Lu  D, Schreiber  NA,  et al.  Association of AR-V7 on circulating tumor cells as a treatment-specific biomarker with outcomes and survival in castration-resistant prostate cancer.  JAMA Oncol. 2016;2(11):1441-1449.PubMedGoogle ScholarCrossref
Original Investigation
February 2018

Positron Emission Tomography/Computed Tomography–Based Assessments of Androgen Receptor Expression and Glycolytic Activity as a Prognostic Biomarker for Metastatic Castration-Resistant Prostate Cancer

Author Affiliations
  • 1Department of Radiology, Memorial Sloan Kettering Cancer Center, New York, New York
  • 2Department of Nuclear Medicine, Centre Chirurgical Marie Lannelongue, Le Plessis Robinson, France
  • 3Department of Medical Physics, Memorial Sloan Kettering Cancer Center, New York, New York
  • 4Department of Epidemiology and Biostatistics, Memorial Sloan Kettering Cancer Center, New York, New York
  • 5Department of Pathology, Memorial Sloan Kettering Cancer Center, New York, New York
  • 6Human Oncology Pathogenesis Program, Memorial Sloan Kettering Cancer Center, New York, New York
  • 7Department of Medicine, Memorial Sloan Kettering Cancer Center, New York, New York
  • 8Department of Medicine, Weill Cornell Medicine, New York, New York
JAMA Oncol. 2018;4(2):217-224. doi:10.1001/jamaoncol.2017.3588
Key Points

Question  Would molecular imaging with positron emission tomography for androgen receptors using fluorodihydrotestosterone F 18 and glycolysis using 2-fluoro-2-D-deoxyglucose F 18 identify biochemical characteristics of castration-resistant prostate cancer that would be predictive for prognosis?

Findings  In a cohort study of 133 patients, clinically relevant heterogeneity affecting prognosis was found in imaging phenotypes for androgen receptor expression and glycolysis on a lesion-by-lesion and individual patient basis.

Meaning  Prognosis of patients with metastatic castration-resistant prostate cancer is adversely affected by easily observed positron emission tomography–based molecular imaging biomarkers (eg, total lesion numbers detected per patient, high 2-fluoro-2-D-deoxyglucose F 18 uptake of individual lesions, and the pattern of androgen receptor and glycolysis lesion phenotypes) within individual patients.

Abstract

Importance  Androgen receptor–signaling inhibitor (ARSi) drugs prolong life in metastatic castration–resistant prostate cancer (mCRPC), but such tumors eventually become resistant and progress. Comprehensive positron emission tomography/computed tomography (PET/CT) imaging using fluoro-2-D-deoxyglucose F 18 ([18F]-FDG) for glycolysis (Glyc) and fluorodihydrotestosterone F 18 ([18F]-FDHT) for androgen receptor (AR) expression determine heterogeneity of imaging phenotypes, which may be useful in distinguishing patients who will benefit from ARSi drugs from those who need alternative treatments.

Objective  To test the hypothesis that PET/CT-based assessments of AR expression and glycolytic activity would reveal heterogeneity affecting prognosis.

Design, Setting, and Participants  Between April 6, 2007, and October 4, 2012, patients with mCRPC underwent imaging with both [18F]-FDG and [18F]-FDHT at Memorial Sloan Kettering Cancer Center. The patients were naive to ARSi treatment with enzalutamide or abiraterone acetate and were referred during documented disease progression. Image-directed biopsy determined the presence or absence of prostate cancer at positive imaging sites.

Interventions  PET/CT imaging was performed with [18F]-FDHT and [18F]-FDG; select individual lesions were biopsied to correlate imaging phenotype with histologic findings.

Main Outcomes and Measures  All metabolically active lesions were interpreted as [18F]-FDHT-positive (AR1) or [18F]-FDHT-negative (AR0) and as [18F]-FDG-positive (Glyc1) or [18F]-FDG-negative (Glyc0). Correlation was performed with overall survival for both individual lesion imaging phenotype as well as patient-specific imaging phenotype.

Results  The mean (SD) age of the 133 patients was 68 (8.6) years. Imaging phenotypes of 2405 PET/CT-positive lesions (median, 12.0 per patient) included 1713 (71.2%) AR1Glyc1, 386 (16.0%) AR1Glyc0, and 306 (12.7%) AR0Glyc1. On multivariate analysis, each phenotype had an independent negative impact effect on survival, most pronounced for AR0Glyc1 lesions (hazard ratio [HR], 1.11; 95% CI, 1.05-1.16; P< .001), followed by AR1Glyc1 lesions (HR, 1.05; 95% CI, 1.03-1.06; P< .001) and AR1Glyc0 lesions (HR, 1.03; 95% CI, 1.00-1.05; P = .048). When sorted by lesion type, 4 patient-specific groups emerged: (1) concordant, with all AR1Glyc1 (34 patients [25.6%]); (2) AR predominant, with AR1Glyc1 and varying numbers of AR1Glyc0 (33 [24.8%]); (3) Glyc predominant, with AR1Glyc1 and varying numbers of AR0Glyc1 (40 [30.1%]); and (4) mixed, with AR1Glyc1 plus a mixture of varying numbers of AR1Glyc0 and AR0Glyc1 (26 [19.5%]).

Conclusions and Relevance  Heterogeneity of PET/CT imaging phenotype has clinical relevance on a lesion and individual patient level. With regard to mCRPC lesions, most express ARs, consistent with initial benefit of ARSi drugs. On a patient basis, 49% (groups 3 and 4) had at least 1 AR0Glyc1 lesion—the imaging phenotype with the most negative effect on survival, possibly due to ARSi resistance.

Introduction

An estimated 26 730 men in the United States will die from prostate cancer in 2017.1 The lethal form of the disease generally occurs when prostate cancer reaches a state of castration resistance and widely metastasizes. The androgen receptor (AR) axis remains functional in this advanced state by several mechanisms, including mutation, overexpression, and ligand-independent activation.2 Effective new drugs that target the AR-signaling axis are in development. Two such agents, abiraterone acetate and enzalutamide, have been approved for treating advanced prostate cancer based on demonstration of improved survival.3-6

As AR-directed treatments become more widely available, improved biomarkers of AR axis activity are needed to optimize patient selection, tumor detection, and monitoring of treatment response and/or progression. We have focused on molecular imaging of men with metastatic castration-resistant prostate cancer (mCRPC), as described by Scher and colleagues.7 We report baseline imaging characteristics of 133 patients with mCRPC who were examined with dual positron emission tomography/computed tomography (PET/CT) imaging using 2-flouro-2-deoxy-D-glucose F 18 ([18F]-FDG) as an indicator of tumor glycolysis (Warburg effect7) and fluorodihydrotestosterone F 18 ([18F]-FDHT), an AR probe. We hypothesized that the imaging phenotypes for these radiotracers would be clinically relevant on both a lesion and patient basis by providing insights into the heterogeneous biology of mCRPC that identify patient subsets with distinct prognostic features for better treatment selection and monitoring.

Methods
Patient Selection

Patients with progressive mCRPC were prospectively enrolled to undergo dual [18F]-FDG and [18F]-FDHT PET/CT imaging. The imaging sessions were performed between April 6, 2007, and October 4, 2012. The objective was to evaluate the performance characteristics of [18F]-FDG and [18F]-FDHT. Eligibility criteria included histologically confirmed adenocarcinoma of the prostate and progressive disease based on either a minimum of 3 rising prostate-specific antigen (PSA) levels taken 1 week apart, with the last result being at least 2 ng/mL or new or progressive soft tissue and/or bone disease confirmed on combined CT/magnetic resonance imaging or bone scan (Figure 1). Patients who had not undergone orchiectomy had to continue androgen deprivation therapy (<50 ng/dL) to maintain castrate levels of testosterone.7 The clinical protocol for this study was reviewed and approved by Memorial Sloan Kettering Cancer Center’s institutional review board. The participants provided written informed consent. There was no financial compensation.

PET/CT Imaging Protocol

PET/CT imaging was performed as previously described8,9 and was temporally scheduled to serve as a baseline prior to entry into contemporary clinical drug trials for AR-directed therapy in which the patients were separately enrolled. [18F]-FDG was obtained from a commercial vendor (IBA Molecular North America Inc) and calibrated by our in-house radiopharmacy immediately before injection of 370 MBq (10 mCi) after a 4- to 6-hour fast (serum glucose levels <200 mg/dL [to convert to millimoles per liter, multiply by 0.0555]). [18F]-FDHT was prepared at Memorial Sloan Kettering Cancer Center as previously described,8,9 and approximately 350 MBq (9 mCi) was injected without dietary preparation. The 2 radiotracers were administered with an interval to allow for decay of 18F (median interval, −1.0 day; range −29 to 12 days). Scans were acquired on a PET/CT camera (Discovery STE; GE Healthcare) from midskull to upper thigh after a 60- to 80-minute ([18F]-FDG) and 30- to 40-minute ([18F]-FDHT) uptake period. A lesion-by-lesion analysis was performed using PET/CT volume computer-assisted reading workstation (GE Healthcare), according to previously described methodology in which lesions are measured quantitatively with maximum standardized uptake value (SUVmax) and qualitatively (confidence scale, 0-4) with a background correction factor to optimize objectivity and interobserver agreement with respect to the imaging detectability threshold.10 Paired PET/CT data for both [18F]-FDG and [18F]-FDHT were recorded for comparison on a per-lesion and per-patient basis.

Lesion Description

We developed a shorthand notation for PET/CT scan–determined molecular imaging phenotypes of 1 for positive or detected and 0 for negative or undetected. Similar to computer code, these symbols are often used for on/off or present/absent. Thus, because [18F]-FDG uptake is based on glycolysis (Glyc), positive or negative [18F]-FDG uptake was noted as Glyc1 and Glyc0, respectively. Similarly, because [18F]-FDHT uptake is based on AR binding, positive or negative [18F]-FDHT lesions were AR1 and AR0, respectively. Accordingly, 3 lesion phenotypes were defined: AR1Glyc1, AR1Glyc0, and AR0Glyc1.

Patient Description

We grouped patients into 4 categories based on imaging characteristics and phenotypes (Figure 2). Given that all patients had at least 1 AR1/Glyc1 lesion, the 4 groups were (1) patients with lesions that were all avid for both tracers (concordant: AR1Glyc1) (34 patients [25.6%]), (2) patients with concordant lesions (AR1Glyc1) and at least 1 lesion that exclusively took up [18F]-FDHT (AR predominant: AR1Glyc0) (33 [24.8%]), (3) patients with concordant lesions (AR1Glyc1) and at least 1 lesion that exclusively took up [18F]-FDG (Glyc predominant: AR0Glyc1) (40 [30.1%]), and (4) patients with a mixture of concordant lesions, AR-predominant lesions, and Glyc-predominant lesions (mixed) (26 [19.5%]).

Tissue Correlation

A subset of patients underwent image-guided biopsies of presumed bone and soft-tissue lesions. Biopsy sites included AR1Glyc1, AR0Glyc1, and PET/CT-negative sites of clinical concern. Histopathology results were reviewed by a urologic pathologist (S.W.F.). prostate-specific membrane antigen (mAb 3E6, 1:100; Dako) and, more commonly, PSA (pAb, RTU; Ventana Medical Systems) immunohistochemical stains were performed in a Clinical Laboratory Improvement Amendments–approved laboratory and correlated with quantitative measures of baseline [18F]-FDG and [18F]-FDHT PET/CT scans, performed within 30 days before or after biopsy. Matching of the biopsy site and PET/CT site was confirmed by comparing interventional radiology and PET/CT images.

Statistical Analysis

Our goal was to assay imaging phenotype on a lesion-by-lesion and per-patient basis to determine biologic and clinically relevant PET/CT features that could serve to enhance our understanding of mCRPC biology in terms of metastatic distribution, estimate prognosis, improve trial design (patient selection in particular), and possibly estimate response to AR-signaling inhibitor (ARSi) therapy. An important companion goal was to develop and validate methodology that allows for analysis of multiple lesions per patient, which is commonplace in mCRPC. A sub-aim of this process was to optimize the application of information from lesion biopsies to provide estimates of accuracy of AR and glycolysis imaging in this patient group.

Standard statistical approaches were described in the protocol for comparing lesions seen on multiple imaging modalities. The following parameters were correlated with survival: SUVmax for [18F]-FDHT, SUVmax for [18F]-FDG, and number of lesions per patient. These parameters were first used as continuous variables and then as dichotomous variables using median split. In addition, 4 distinct groups were created and tested for significant associations with survival. All associations between categorical variables and survival were evaluated using the log-rank test, and associations between continuous variables and survival were evaluated using Cox proportional hazards regression. Several [18F]-FDG and [18F]-FDHT PET/CT parameters were compared and correlated with standard clinical variables using both graphical methods (eg, scatterplots) and statistical metrics (eg, correlation coefficients). These analyses of agreement were performed on a lesional basis. Prognostic values of imaging parameters were assessed using standard survival analysis methods, such as Kaplan-Meier estimation for survival probabilities, log-rank test for comparison of groups, and Cox proportional hazards regression model for regression analysis.

As data analysis proceeded, we identified a need to develop and apply more specialized statistical techniques that could provide estimates about the positive predictive value of the imaging if we could have biopsied all of the lesions. Accordingly, the expected number of true-positives was calculated using a hierarchical bayesian model—specifically, the β-binomial.11 The results of imaging correlation with biopsy were considered a priori information, and a posteriori calculation was used to determine the best estimate (mean and 95%CI). This method is similar to a previous approach.12 eAppendix 1 and eFigures 1-4 in the Supplement provide more details.

Results

A total of 158 patients were selected during the enrollment period (Figure 1). Of these, 25 patients were excluded. Thus, a total of 133 patients were included (eTable 1 in the Supplement). Median age was 68 years (range, 44-85 years), and mean (SD) was 68 (8.6) years. Median and mean baseline PSA levels were 42.72 and 123.72 ng/mL, respectively (range 0-1477 ng/mL [1:1 conversion to micrograms per liter]). All patients had previously received castration therapy. A total of 68 patients (51.1%) had received prior taxane therapy (eTable 1 in the Supplement).

Lesion Analysis

A total of 2405 metabolically active lesions were detected on the combined [18F]-FDG and [18F]-FDHT PET/CT scan sets (mean, 17.8; median, 12.0; range, 1-61 per patient). Of these lesions, 1997 (83.0%) were in bone, 342 (14.2%) were in lymph nodes, and 66 (2.7%) were in other soft-tissue sites. A total of 116 (87.2%) patients had at least 1 PET/CT-positive bone lesion, including 33 (28.4%) patients with fewer than 5 lesions, 16 (13.8%) with 5 to 9 lesions, 21 (18.1%) with 10 to 19 lesions, 35 (30.2%) with 20 to 39 lesions, and 11 (9.5%) with 40 or more lesions. Seventeen of 133 (12.8%) patients had no bone lesions on PET/CT, and 50 (37.6%) patients harbored only bone lesions. Median [18F]-FDHT SUVmax values were higher than median [18F]-FDG SUVmax values in bone (5.5 vs 3.6), lymph nodes (6.4 vs 3.9), prostate bed (7.6 vs 4.9), and other soft-tissue sites (5.0 vs 4.5) with the exception of the lung (1.7 vs 2.9) (eFigure 5 in the Supplement). The aggregate pool of paired lesional [18F]-FDG and [18F]-FDHT SUVmax values was weakly correlated (R2 = 0.01) (eFigure 6 in the Supplement). Imaging phenotypes of the 2405 PET/CT-positive lesions were: 1713 (71.2%) AR1Glyc1, 386 (16.0%) AR1Glyc0, and 306 (12.7%) AR0Glyc1. Thus, 87.3% of lesions were AR1 and 84.0% were Glyc1 on image analysis.

Patient Imaging Phenotypes

Four patient groups emerged, as described above. The concordant group (median, 5.5 lesions per patient) had significantly fewer lesions than all other groups (median of all other groups, 19.0; P < .001) (eTable 2 in the Supplement). Sixty-six (49.6%) patients had at least 1 lesion with absent or altered AR (AR0Glyc1), including a median of 3.5 AR0Glyc1 lesions per patient in the Glyc-predominant group and 2.0 lesions per patient in the mixed group. Baseline PSA expression was greatest for patients in the AR-predominant group, with a median level of 56.4 ng/mL compared with 21.2 ng/mL for the Glyc-predominant group (P = .02) (eFigure 7 in the Supplement).

Overall Survival

Median patient survival was 91.6 weeks (5-year survival, 11.0%), with a median follow-up for censored patients of 264.6 weeks (range, 56-433 weeks). Patients with more than the median of 12 metabolizing lesions had a significantly worse overall survival (OS) than patients with fewer than 12 lesions (hazard ratio, [HR], 3.01; 95% CI, 2.04-4.46; P < .001), as did patients with median [18F]-FDG SUVmax greater than 7.6 (HR, 1.86; 95% CI, 1.29-2.69; P < .001). Dichotomizing patients with median [18F]-FDHT SUVmax of 9.4 (P = .07) did not yield a significant survival advantage (Figure 3). After excluding AR0Glyc1 lesions, OS remained poor for patients with high AR1 lesion counts (n = ≥10) (67.9 vs 175.6 weeks; P < .001) (eFigure 8 in the Supplement). On multivariate analysis using a Cox proportional hazards regression model, the 3 imaging phenotypes were found to be independent of each other and statistically significant when correlated with reduced survival. On multivariate analysis, the results indicated that each additional AR0Glyc1 lesion increases the risk of death by 11% (HR, 1.11; 95% CI, 1.05-1.16; P< .001), AR1Glyc1 lesion by 5% (HR, 1.05; 95% CI, 1.03-1.06; P< .001), and AR1Glyc0 lesion by 3% (HR, 1.03; 95% CI, 1.00-1.05; P = .048) when the number of the other 2 lesion phenotypes is held constant. Baseline PSA level was not prognostic of OS (P = .06) (eFigure 7 in the Supplement).

Biopsy Data

To better understand the biologic significance of the positive images on a lesion-by-lesion basis, we obtained biopsies for confirmation of pathologic findings in 50 patients (37.6%) (eFigure 9 in the Supplement). A total of 59 biopsies were obtained (33 bone, 26 soft tissue), 2 of which were nondiagnostic. The remaining 57 biopsies were interpreted as follows: prostate cancer, 48 (84.2%); nonprostatic cancer, 3 (5.3%); inflammation, 1 (1.8%); and normal or benign tissue, 5 (8.8%). All 42 AR1 lesions were prostate cancer, but 6 of 48 (12.5%) biopsies that were positive for prostate cancer were AR0Glyc1. Of 3 biopsies yielding a nonprostatic primary cancer, all were AR0Glyc1. One biopsy yielding inflammation was AR0Glyc1. Of the 5 biopsies reported as benign, 4 were Glyc nondetectable and all 5 were AR nondetectable (eTable 3 in the Supplement). A summary of the scan findings for bone and soft tissue is reported in eTable 4 in the Supplement. Twelve biopsy specimens were interpreted as consistent with poorly differentiated prostate cancer for which PSA immunohistochemistry was performed; 7 of the patient samples were AR1 and 5 were AR0. All 7 of the AR1 scans were PSA positive, and 1 of 5 of the AR0 scans were PSA positive. χ2 Analysis indicated a significant association between AR1 and PSA expression, a marker of differentiation in prostate cancer (P = .003 for chance association).

Biopsy Results as A Priori Predictors of Best Estimates for In Vivo Metabolic Heterogeneity

Because we can only biopsy a few of the image-positive sites, we assessed the accuracy of PET/CT imaging in statistical terms by applying a hierarchical statistical model based on Bayes’ rule to predict the true-positive results for PET/CT imaging. For example, in the case of [18F]-FDHT lesion uptake, all 42 of the AR1 scan sites were confirmed by biopsy to be prostate cancer. However, 2057 positive scan sites were not biopsied, which raises the question: What is the actual number of true-positive sites likely to be? The model provides a best estimate of 98% of the 2057 unbiopsied lesions (2009), and adding the biopsy-confirmed sites (2009 + 42 = 2051) provides a best estimate of the total number of lesions in the AR1 lesion group that are prostate cancer (eTable 4 in the Supplement). We used the best estimates computed as described above to correctly obtain the true a posteriori proportion of prostate cancer–positive lesions that corresponded with each of the 3 lesion phenotype combinations for the 2 metabolic tracers: (1) concordant, approximately 83% AR1/Glyc1; (2) AR only, approximately 13% AR1/Glyc0; and (3) Glyc only, approximately 4% AR0/Glyc1 (eFigure 4 in the Supplement). After bayesian correction, fully 96% of the lesions were AR1, providing a biologic rationale for treatment response to AR inhibition and supporting a major role for ARs in initiating mCRPC.

Discussion

Prostate cancer is a heterogeneous disease throughout its natural history, even in its most advanced stage, mCRPC. Clinical features, such as Gleason score, hemoglobin level, and lactate dehydrogenase, can prognosticate for OS and distinguish between higher- and lower-risk patients.13 Furthermore, as patients develop mCRPC, biopsies of individual sites of disease demonstrate increasing genomic mutational burden,14 as does the heterogeneity of circulating tumor cells.15

In this study, we demonstrated intrapatient lesional diversity using molecular imaging of the primary driver of prostate cancer growth (ie, the AR), as well as Glyc, and showed that the 2 biochemical features appear to be largely independent of one another. We introduced unique features of analysis and experience, including comparison of [18F]-FDG and [18F]-FDHT, in an mCRPC patient cohort under specific protocols (eAppendix 2 and eFigure 10 in the Supplement provide specific information on PET/CT volume computer-assisted reading and [18F]-FDG and [18F]-FDHT radiopharmacology that were essential to data interpretation). The results yielded 4 patient groups based on patterns of lesional diversity. Those whose lesions demonstrated concordant [18F]-FDHT and [18F]-FDG uptake were found to have the best survival. Those whose disease manifested with a preponderance of [18F]-FDHT-null lesions, presumably due to a paucity of overexpressed wild-type AR, had the poorest prognosis. In terms of overall number of lesions per patient, OS in patients with 12 or more lesions (median number per patient) was sharply lower than in those with fewer lesions (HR, 3.2; P < .001). Patients who harbored at least 1 glyc-detectable lesion with an SUVmax greater than 7.6 (median of all [18F]-FDG values) also had reduced OS (HR, 1.9; P < .001) (Figure 3).

We found that [18F]-FDHT positivity (AR1) was a highly specific indicator for histologic findings of prostate cancer (98% predicted true-positive rate). AR0 lesions were seen in few prostate cancers, and based on correlation with PSA expression, these lesions were less well differentiated. However, [18F]-FDG was less specific (5 [45.5%] of AR0Glyc1 lesions were nonprostate cancer).

The use of [18F]-FDHT and [18F]-FDG imaging to appreciate intrapatient lesional heterogeneity has considerable practical applications for drug development in prostate cancer. Precision medicine benefits from knowledge about the pathology and genome of individual lesions, which may call for specific drugs or modalities for effective treatment.14 But heterogeneity has been the primary vulnerability of targeted therapy, and biopsying all lesions in a given patient is a practical impossibility. One possible solution is for imaging phenotypes to direct a biopsy to a particular lesion, which may provide specifics on how to treat a patient for improved response. As expected, the most common phenotype, AR1Glyc1, and the variant AR1Glyc0, together make up 96% of active mCRPC lesions. This finding is consistent with the importance of ARs to the development of mCRPC and as a target for the effectiveness of ARSi drugs. We hypothesize that, among the lesion-imaging phenotypes we observed, the AR0Glyc1 phenotype should provide clinically relevant information. Although producing a relatively small number of lesions overall when considered on a patient distribution basis, the AR0Glyc1 phenotype represents 1 or more lesions in nearly half of patients. Since we hypothesize that AR0Glyc1 lesions are likely to have altered or absent ARs, we have planned follow-up studies to evaluate the response to second-generation ARSi drugs in patients expressing this phenotype. Furthermore, the AR0Glyc1 subset contains nonprostate cancer pathologies. Both lesion histologic factors will require additional alternative targeted therapies.

Because patients may have either intrinsic or acquired resistance to ARSi drugs over a few months of treatment, a search for alternative targets and supplementary targeted therapies is under way. For example, a recent review of biopsy findings in 150 individuals with mCRPC revealed a higher-than-anticipated mutational burden, raising the possibility of new treatment paradigms, including an array of non–AR-blocking targeted therapies.14

The present study will serve as the baseline for future studies to determine whether individual patient groups or the presence of individual lesion subtypes may have distinct patterns of treatment response and/or survival; for example, the lack of detectable ARs (AR0Glyc1) may correlate with poor response to ARSi drugs. Finally, we recognize that imaging is only 1 feature of prognostic modeling in prostate cancer. Separately, we are analyzing [18F]-FDG and [18F]-FDHT data as they relate to other known prognostic determinants for OS in multivariate models.

Limitations

We recognize that this study has potential limitations. Among the more important of these, only 2 molecular imaging tracers were used, namely, [18F]-FDG and [18F]-FDHT, and although these tracers likely detected most lesions, it is possible that other active lesions were missed. In addition, lesions outside the PET/CT field of view (ie, above midskull or below upper thigh) may have been missed. Finally, in terms of tissue correlation, only a subset of lesions could practically be biopsied. This factor may have introduced a bias related to lesion accessibility, which limits generalization of the findings to all lesions.

Conclusions

In mCRPC, PET/CT imaging with [18F]-FDG and [18F]-FDHT is clinically feasible before treatment with second-generation ARSi drugs or chemotherapy. Disease burden, as represented by lesion number and [18F]-FDG SUV max greater than 7.6 as an indicator of Warburg effect, are powerful prognostic biomarkers. Bayesian analysis of image-directed biopsies predicted best estimate +/− 95% CIs for 3 individual mCRPC phenotypes: approximately 83.1% (range, 81.1%-85.0%) were AR1Glyc1, approximately 13.4% (range, 12.0%-14.6%) were AR1Glyc0, and approximately 3.5% (range, 1.8%-5.2%) were AR0Glyc1 (eFigure 9 in the Supplement). Each phenotype individually contributes to reduced survival. The most potent of these phenotypes with respect to adverse prognosis is AR0Glyc1. Such lesions should be considered for biopsy because nonprostatic cancer may be significantly represented in this subgroup; for prostate cancer lesions, clinically actionable mutations may be identified.

Back to top
Article Information

Accepted for Publication: August 18, 2017.

Corresponding Author: Steven M. Larson, MD, Department of Radiology, Memorial Sloan Kettering Cancer Center, 417 E 68th St, New York, NY 10065 (larsons@mskcc.org).

Published Online: November 9, 2017. doi:10.1001/jamaoncol.2017.3588

Author Contributions: Drs Morris and Larson contributed equally to the study, had full access to all the data in the study, and take responsibility for the integrity of the data and the accuracy of the data analysis.

Study concept and design: Fox, Gönen, Vargas, Schöder, Sawyers, Weber, Scher, Morris, Larson.

Acquisition, analysis, or interpretation of data: Fox, Gavane, Blanc-Autran, Nehmeh, Gönen, Beattie, Schöder, Humm, Fine, Lewis, Solomon, Osborne, Scher, Morris, Larson, Veach.

Drafting of the manuscript: Fox, Gavane, Gönen, Fine, Lewis, Weber, Morris, Larson.

Critical revision of the manuscript for important intellectual content: Fox, Blanc-Autran, Nehmeh, Beattie, Vargas, Schöder, Humm, Fine, Solomon, Osborne, Sawyers, Scher, Morris, Larson, Veach.

Statistical analysis: Fox, Gönen.

Obtained funding: Lewis, Scher, Larson.

Administrative, technical, or material support: Blanc-Autran, Beattie, Humm, Fine, Lewis, Scher, Morris.

Study supervision: Fox, Blanc-Autran, Schöder, Fine, Lewis, Sawyers, Weber, Morris, Larson.

Conflict of Interest Disclosures: Dr Solomon receives research support from GE Healthcare, serves as a consultant with Johnson & Johnson and AstraZeneca, and is a shareholder of Progenics. Dr Sawyers is a co-inventor of enzalutamide and entitled to royalties and serves on the Board of Directors of Novartis. Dr Scher receives personal fees from Astellas, BIND Pharmaceuticals, Blue Earth Diagnostics, Clovis Oncology, Merck, Sanofi, WIRB Copernicus Group, Asterias Biotherapeutics, Physicians Education Resource, and OncLive Insights; nonfinancial support from Bristol-Myers Squibb, Ferring Pharmaceuticals, Medivation, Millennium, Exelixis, and Janssen Research and Development; and grants from Illumina Inc, Innocrin Pharma, and Janssen. Dr Morris is an uncompensated consultant to Astellas, Bayer, and Endocyte and a compensated consultant for Tokai and AAA; he also receives research support through his institution for the conduct of clinical trials from Progenics, Endocyte, and Bayer. No other disclosures are reported.

Funding/Support: The study was supported by Cancer Center Support grant P30 CA008748 from the National Institutes of Health/National Cancer Institute, grant P50 CA086438 from the In Vivo Cellular and Molecular Imaging Center, the prostate cancer program of Memorial Sloan Kettering Cancer Center, the Center for Targeted Radioimmunotherapy and Theranostics of the Ludwig Center for Cancer Immunotherapy, and the David H. Koch Foundation. The Radiochemistry & Molecular Imaging Probes Core of Memorial Sloan Kettering Cancer Center is also supported in part by Cancer Center Support Grant P30 CA008748.

Role of the Funder/Sponsor: The funding organizations had no role in the design and conduct of the study; collection, management, analysis, and interpretation of the data; preparation, review, or approval of the manuscript; and decision to submit the manuscript for publication.

Additional Contributions: Leah Bassity, MA (Memorial Sloan Kettering Cancer Center), provided editorial assistance; there was no financial compensation. We thank the research study assistants, clinical research coordinators, research managers, research nurses, technologists, and radiopharmacists for making this study possible.

References
1.
Siegel  RL, Miller  KD, Jemal  A.  Cancer statistics, 2017.  CA Cancer J Clin. 2017;67(1):7-30.PubMedGoogle ScholarCrossref
2.
Chen  Y, Sawyers  CL, Scher  HI.  Targeting the androgen receptor pathway in prostate cancer.  Curr Opin Pharmacol. 2008;8(4):440-448.PubMedGoogle ScholarCrossref
3.
de Bono  JS, Logothetis  CJ, Molina  A,  et al; COU-AA-301 Investigators.  Abiraterone and increased survival in metastatic prostate cancer.  N Engl J Med. 2011;364(21):1995-2005.PubMedGoogle ScholarCrossref
4.
Fizazi  K, Scher  HI, Molina  A,  et al; COU-AA-301 Investigators.  Abiraterone acetate for treatment of metastatic castration-resistant prostate cancer: final overall survival analysis of the COU-AA-301 randomised, double-blind, placebo-controlled phase 3 study.  Lancet Oncol. 2012;13(10):983-992.PubMedGoogle ScholarCrossref
5.
Scher  HI, Fizazi  K, Saad  F,  et al; AFFIRM Investigators.  Increased survival with enzalutamide in prostate cancer after chemotherapy.  N Engl J Med. 2012;367(13):1187-1197.PubMedGoogle ScholarCrossref
6.
Beer  TM, Armstrong  AJ, Rathkopf  DE,  et al; PREVAIL Investigators.  Enzalutamide in metastatic prostate cancer before chemotherapy.  N Engl J Med. 2014;371(5):424-433.PubMedGoogle ScholarCrossref
7.
Scher  HI, Morris  MJ, Basch  E, Heller  G.  End points and outcomes in castration-resistant prostate cancer: from clinical trials to clinical practice.  J Clin Oncol. 2011;29(27):3695-3704.PubMedGoogle ScholarCrossref
8.
Larson  SM, Morris  M, Gunther  I,  et al.  Tumor localization of 16β- 18F-fluoro-5α-dihydrotestosterone versus 18F-FDG  in patients with progressive, metastatic prostate cancer.  J Nucl Med. 2004;45(3):366-373.PubMedGoogle Scholar
9.
Zanzonico  PB, Finn  R, Pentlow  KS,  et al.  PET-based radiation dosimetry in man of 18F-fluorodihydrotestosterone, a new radiotracer for imaging prostate cancer.  J Nucl Med. 2004;45(11):1966-1971.PubMedGoogle Scholar
10.
Fox  JJ, Autran-Blanc  E, Morris  MJ,  et al.  Practical approach for comparative analysis of multilesion molecular imaging using a semiautomated program for PET/CT.  J Nucl Med. 2011;52(11):1727-1732.PubMedGoogle ScholarCrossref
11.
Gönen  M, Schrag  D, Weiser  MR.  Nodal staging score: a tool to assess adequate staging of node-negative colon cancer.  J Clin Oncol. 2009;27(36):6166-6171.PubMedGoogle ScholarCrossref
12.
Pandit-Taskar  N, O’Donoghue  JA, Durack  JC,  et al.  A phase I/II study for analytic validation of 89Zr-J591 immunoPET as a molecular imaging agent for metastatic prostate cancer.  Clin Cancer Res. 2015;21(23):5277-5285.PubMedGoogle ScholarCrossref
13.
Halabi  S, Lin  CY, Kelly  WK,  et al.  Updated prognostic model for predicting overall survival in first-line chemotherapy for patients with metastatic castration-resistant prostate cancer.  J Clin Oncol. 2014;32(7):671-677.PubMedGoogle ScholarCrossref
14.
Robinson  D, Van Allen  EM, Wu  YM,  et al.  Integrative clinical genomics of advanced prostate cancer.  Cell. 2015;161(5):1215-1228.PubMedGoogle ScholarCrossref
15.
Scher  HI, Lu  D, Schreiber  NA,  et al.  Association of AR-V7 on circulating tumor cells as a treatment-specific biomarker with outcomes and survival in castration-resistant prostate cancer.  JAMA Oncol. 2016;2(11):1441-1449.PubMedGoogle ScholarCrossref
×