Novel Framework for Treatment Response Evaluation Using PSMA PET/CT in Patients with Metastatic Castration-Resistant Prostate Cancer (RECIP 1.0): An International Multicenter Study

Visual Abstract

Inmet astatic prostate cancer, treatment response is typically evaluated using conventional imaging (CT and bone scanning) according to the Prostate Cancer Working Group Criteria 3 (PCWG3) guidelines. Prostate-specific membrane antigen (PSMA)-targeted PET/CT is a novel imaging technique that showed greater detection accuracy than conventional imaging in patients with high-risk primary prostate cancer (1). The U.S. Food and Drug Administration approved [ 68 Ga]Ga-PSMA-11 PET/CT for different clinical settings in men with prostate cancer (2). However, there is little evidence for the prognostic value of PSMA PET/CT for response assessment in men with advanced prostate cancer (3,4). In our clinical experience using PSMA PET/CT for response evaluation of systemic metastatic castration-specific prostate cancer (mCRPC) treatments, a decrease in total disease burden can coincide with appearance of new lesions. This scenario, referred to as heterogeneous response, often leaves the treating physician in a clinical dilemma (5). Considering the rapidly evolving era of targeted treatments for mCRPC, accurate and early response assessment is urgently needed, but standardized response evaluation criteria for PSMA PET imaging have not been developed yet.
[ 177 Lu]Lu-PSMA ( 177 Lu-PSMA) is a small-molecule inhibitor that binds with high affinity to PSMA and delivers b-radiation.
The randomized TheraP trial demonstrated superior prostatespecific antigen (PSA) responses and progression-free survival for 177 Lu-PSMA-617 versus cabazitaxel (6). In the phase III VISION trial, 177 Lu-PSMA-617 prolonged overall survival (OS) and imaging-based progression-free survival, when added to the standard of care in patients with metastatic castration-resistant prostate cancer (mCRPC) (7).
This study had 2 key objectives: first, to develop version 1.0 of a standardized framework for response evaluation criteria in PSMA PET/CT (RECIP) in men with mCRPC who undergo 177 Lu-PSMA and, second, to develop a composite response classification that combines PSA measurements and PSMA PET/CT responses by RECIP 1.0 (PSA 1 RECIP).

Patients and Study Design
In this international multicenter study, men with mCRPC treated with 177 Lu-PSMA-I&T or 177 Lu-PSMA-617 between December 10, 2014, and July 19, 2019, at the Technical University Munich, UCLA, and University Hospital Essen were retrospectively screened for inclusion. Eligible patients had received PSMA PET/CT at baseline (bPET) and after 2 cycles of treatment (interim PET/CT [iPET]), had received the same PET radiotracer at bPET and iPET, and had survival data available. 177 Lu-PSMA was administered by intravenous injection of 6.0-8.5 GBq at 6-to 8-wk intervals. Treatment was continued up to a maximum of 4 or 6 cycles in the absence of progression and lack of severe toxicity according to the treating physician. bPET was performed within 10 wk before treatment. iPET was performed at 12 6 2 wk after treatment initiation and 5 6 1 wk after the second treatment cycle. Treatment protocols are detailed in the supplemental materials (available at http://jnm. snmjournals.org) (8)(9)(10)(11)(12)(13). Serum PSA measurements were also collected at baseline and at 12 6 2 wk. Changes in PSA levels at 12 wk relative to baseline were recorded and categorized according to PCWG3 criteria as response ($50% decrease) or progression ($25% increase) (14).
The primary outcome measure was the prognostic value of RECIP 1.0 for OS. The secondary outcome measure was the prognostic ability of PSA 1 RECIP versus PSA only (Fig. 1).
All patients gave written informed consent to undergo clinical PSMA PET/CT. The retrospective analysis was approved by the Ethics Committees of each participating site (Technical University Munich, approval 115/18S; UCLA, approval 20-000954, University Hospital Essen, approval 19-8570-BO), and the committees waived the necessity for study-specific consent. Of note, the patient population in this study to develop RECIP was used to compare different criteria for response assessment in mCRPC (15).

Imaging Acquisition
Images were obtained after application of PSMA ligands that were synthesized as described previously (16,17). Patients received an average (6SD) of 126 6 4 and 317 6 9 MBq of [ 68 Ga]Ga-PSMA-11 and [ 18 F]rhPSMA-7/7.3, respectively, via intravenous bolus. Image acquisition began 71 6 6 min after tracer injection. Data from the CT scan were used for attenuation correction. Images were acquired using Siemens Biograph mCT (n 5 115) and Siemens Biograph 64 (n 5 9) scanners. All images were obtained in accordance with the European Association of Nuclear Medicine guidelines (E-PSMA) for treatment monitoring in patients with mCRPC, ensuring harmonized quantification (18). Standard vendor-provided image reconstructions were used. The institutional applied reconstruction parameters are summarized in Supplemental Table 1. Paired bPET and iPET were performed using the same PET/CT scanner and following same image reconstruction protocol.

Image Analysis
PET/CT datasets from each participating site were anonymized and centralized.
New Lesions. Pairs of bPET and iPET scans were read independently by 3 nuclear medicine physicians, who were masked to outcome data and were not involved in study design. Each reader was provided with full anonymized PET/CT datasets and was asked to assess the scans for new lesions following predefined criteria (Table 1). Disagreement among readers was solved in consensus sessions. Responses in PSMA-VOL were tested in conjunction with appearance of new lesions for associations with OS. We hypothesized, first, that patients with PSMA-VOL_PR without new lesions have OS superior to that of patients with PSMA-VOL_PR and new lesions and, second, that patients with PSMA-VOL_PD and new lesions have worse OS than patients with PSMA-VOL_PD without new lesions. On the basis of our hypothesis, RECIP 1.0 was developed and designed to classify patients into 4 categories: complete response (RECIP-CR), partial response (RECIP-PR), progressive disease (RECIP-PD), and stable disease (RECIP-SD) ( Table 1). Associations of RECIP responses on iPET with OS were evaluated. Further, RECIP responses on iPET were combined with PSA responses at 12 wk to develop a novel composite response classification (PSA 1 RECIP). Definitions of all 3 response classifications are given in Table 1. The prognostic ability of PSA, RECIP, and PSA 1 RECIP responses for OS was evaluated.

Statistical Analysis
Values are reported as average and SD or median and interquartile range (IQR) for continuous variables and as number and percentage for categoric variables. OS was estimated using the Kaplan-Meier method. The associations between OS and appearance of new lesions, changes in PSMA-VOL, and RECIP were evaluated using univariate Cox regression analyses. The hazard ratio (HR), its 95% CI, and the corresponding P values were derived. Appearance of new lesions and PSMA-VOL were tested separately and in combination to identify combined criteria with highest associations with OS. The prognostic ability of the PSA, RECIP, and PSA 1 RECIP classification systems was assessed using the Harrell concordance index (C-index) (20). Comparisons (P values) of C-indices were computed using the concordance function, which estimates the variancecovariance matrix between the correlated (repeated measure) C-indices (21). Agreement between readers in identifying new lesions on iPET was evaluated by Fleiss k (KappaM package) (22). Analyses were performed using R software, version 3.4. A P value of less than 0.05 was considered statistically significant.

RESULTS
From October 1, 2019, to December 18, 2019, retrospective data from 287 men with mCRPC were screened. Of these, 124 (43%) met the eligibility criteria and were included (Consolidated Standard of Reporting Trials diagram; Supplemental Fig. 2). One hundred fifteen (93%) of 124 patients were treated under compassionate-access programs, whereas 9 (7%) were enrolled in a phase II clinical trial (NCT03042312). Baseline characteristics are summarized in Table 2   The median change in PSMA-VOL on iPET relative to bPET was 22.2% (IQR, 239.8 to 146.2). The C-indices for each cut point for definition of response and progression are provided in Supplemental Table 2. A cutoff of 120% had the highest prognostic value for PSMA-VOL_PD with OS (C-index, 0.64). Cutoffs of 220% and 230% had the highest but similar prognostic value for PSMA-VOL_PR with OS (C-index, 0.62), and the 230% cutoff  was chosen to minimize the impact of measurement errors or biologic variability. Stable disease (PSMA-VOL_SD) was defined as either less than a 30% decrease or less than a 20% increase in PSMA-VOL.

Establishment of RECIP and Associations with Overall Survival
RECIP-CR. Absence of any PSMA-ligand uptake on iPET was not observed.
RECIP-PR. Men with PSMA-VOL_PR and no evidence of new lesions had OS superior to that of men with PSMA-VOL_PR and appearance of new lesions (HR, 0.50; 95% CI, 0.25-0.93; P 5 0.039). On this basis, the definition of RECIP-PR was maintained.
RECIP-PD. Men with PSMA-VOL_PD and appearance of new lesions had OS inferior to that of men with PSMA-VOL_PD but no evidence of new lesions (HR, 4.50; 95% CI, 1.36-14.90; P 5 0.014) (Supplemental Fig. 4). On this basis, the definition of RECIP-PD was maintained. A case example of a patient with RECIP-SD is presented in Supplemental Figure 5.

DISCUSSION
Currently, the efficacy of 177 Lu-PSMA and other systemic treatments of mCRPC is evaluated using conventional imaging (bone scanning 1 CT by PCWG3 criteria (14)), which may not accurately assess responses, especially for bone metastases, which are present in about 90% of mCRPC patients. PSMA PET/CT demonstrated a higher detection rate than conventional imaging (1); however, its prognostic role for treatment monitoring has not been established. Criteria for monitoring tumor response in PET imaging were described previously for 18    We developed RECIP 1.0 as the first-to our knowledgeevidence-based framework for response evaluation in prostate cancer using PSMA PET imaging. Two criteria have previously been proposed for the same purpose; however, these proposals included clinical information and were not based on multicenter validation (24,25). Compared with PERCIST, which uses measurements of individual lesions, RECIP 1.0 quantifies changes in total tumor volume, capturing the entire extent of disease. Binary PCWG3 classifies into progressive disease versus nonprogressive disease but lacks the ability to capture response by subcategorizing nonprogressive disease into complete response, partial response, or stable disease. Although identification of progressors may suffice in clinical practice, the objective response rate is commonly used in clinical trials as an endpoint to determining a drug's efficacy (26). To enable assessment of the objective response rate of tumors, RECIP 1.0 was designed to distinguish true responders from patients with stable disease. A heterogeneous response by individual metastatic lesions is quite common during treatment of advanced mCRPC (5), as was confirmed in our patient population; that is, 13% of the patients had new lesions despite a response in tumor burden. These patients were classified by RECIP 1.0 as having stable disease and had a survival outcome different from true responders, who have no sign of progression (i.e., appearance of new lesions), and true progressors, who have both an increase in tumor burden and appearance of new lesions (median OS, 13.1 vs. 21.7 vs. 8.3 mo, respectively).
PSA response ($50% decrease) is commonly used in phase II clinical trials of mCRPC as a primary endpoint to estimate antitumor activity. Our composite response classification system (PSA 1 RECIP) showed a prognostic accuracy for OS superior to that of PSA measurements only, highlighting the potential benefit of combining PSA and RECIP responses into a composite efficacy endpoint for clinical trials of mCRPC. The advantages of using composite endpoints include greater statistical precision and efficiency, that is, smaller sample sizes (which enable less costly trials and lower rates of treatment-related side effects) and a shorter follow-up (which enables faster availability of the results). Nevertheless, designing and implementing such endpoints can be challenging and hence require caution (27). In comparison to PSA measurements, PSMA PET/CT offers additional information about metastatic site and pattern of spread, as well as potential bone complications (e.g., spinal cord compression or fractures). This consideration is highly relevant to the clinical management of patients when adjuvant treatments can be considered, such as emergency surgery, radiation, or other metastasis-directed therapies.
Nomograms to predict outcome after 177 Lu-PSMA using baseline patient and tumor characteristics were developed previously (28). The number of PSMA-positive metastases on pretherapeutic PSMA PET/CT was used as a surrogate marker of tumor volume for easier clinical implementation. Notably, the present analysis investigated dynamic changes in tumor burden during treatment. In this setting, changes in number of lesions are of limited use, and quantitative measurements of tumor burden are essential for accurate response evaluation.
Clinical use of PSMA PET/CT often lacks the ability to quantify whole-body disease burden because of high disease burden in metastatic settings. To enable quantitative assessment of total disease burden during treatment, different vendors are currently developing software tools. For this retrospective study, we used qPSMA for semiautomatic extraction of total tumor volume (19). It is in-housedeveloped and is freely available for widespread use. Other types of dedicated segmentation software might also become available to enable clinical implementation of PSMA PET/CT as a quantitative imaging biomarker in practice and trials (29)(30)(31). The prognostic value of iPET by RECIP is optimal (C-index, 0.65-0.70). The limited prognostic value of RECIP might be caused by an artificial decrease in PSMA expression because of dedifferentiation and not by a true decrease in tumor size.
The major limitations of this study were the lack of a prospective validation of RECIP criteria and an external validation of their threshold definition. Repeatability thresholds for tumor SUV measurements for 68 Ga-PSMA-11 PET/CT were determined previously; however, tumor volumes were not included in the analysis (32). Another limitation of the study is that we could report only the prognostic, not the predictive, value of RECIP since we did not analyze data from a randomized trial powered for outcome. Future randomized studies monitoring tumor response with PSMA PET/ CT are warranted to determine whether higher rates of PSMA response to a drug translate into a better clinical outcome. Also, this study could not compare the prognostic ability of RECIP versus PCWG3 criteria, since bone scans were not included in the clinical workup of 177 Lu-PSMA radionuclide therapy at all institutions. Another limitation is that different PSMA PET radiotracers were used in this study, albeit consistent within patients. Further, the fact that the iPET was performed at 4-6 wk after the second cycle of treatment could impact both the ability to observe disease regression and the opportunity for new lesions to develop. Last, progression-free survival data were not included as a secondary endpoint because clinical assessment was not performed uniformly or at consistent time points across patients. Strengths of the study include the multicentric setting, a large patient population, and long-term follow-up survival data.
Our study has important clinical implications. First, it demonstrates the prognostic role of iPET as a response biomarker to monitor the efficacy of 177 Lu-PSMA and possibly other mCRPC systemic therapies. After the positive outcome of the VISION registration trial (7), approval of 177 Lu-PSMA is imminent. Early and accurate treatment response assessment by PSMA PET/CT may identify nonresponders early in the course of treatment and consequently decrease overtreatment and guide these patients to more effective therapies. Our interim time point of 12 wk for early response evaluation is in line with PCWG2 recommendations for mCRPC and with European Association of Nuclear Medicine procedure guidelines for 177 Lu-PSMA therapy (13,33). End-treatment response evaluation using PSMA PET may also provide useful information on whether patients who complete the maximum number of 177 Lu-PSMA cycles are candidates for a treatment rechallenge (34). However, only a subgroup of patients responds well and completes all cycles (i.e., 39/124 [31%] of our patients), and therefore, such analysis is limited by sample size. Further, there is currently no consensus among specialists on the maximum number of 177 Lu-PSMA cycles. Second, RECIP 1.0 was developed as a potential powerful tool to determine imaging responses and to better assess heterogeneous response, and third, our findings suggest the value of adding PSMA PET/CT imaging to PSA measurements in evaluating treatment efficacy, which may result in higher precision and patient outcome of mCRPC trials. CONCLUSION RECIP 1.0 was developed as an evidence-based novel framework to assess tumor response early in the course of treatment in mCRPC using PSMA PET/CT. PSA 1 RECIP is proposed as a novel composite-efficacy endpoint for clinical trials of mCRPC. PSMA PET/ CT can be used as a response biomarker for early monitoring of the efficacy of 177 Lu-PSMA and potentially other mCRPC treatments. Validation of the findings in a prospective setting is warranted.