Clinical insignificance of [ 18 F ] PSMA ‐ 1007 avid non ‐ specific bone lesions : a retrospective evaluation

Purpose [18F]PSMA-1007 offers advantages of low urinary tracer excretion and theoretical improved spatial resolution for imaging prostate cancer. However, non-specific bone lesions (NSBLs), defined as mild to moderate focal bone uptake without a typical morphological correlate on CT, are a common finding on [18F]PSMA-1007 PET/CT. The purpose of this study was to investigate the clinical outcomes of patients with [18F]PSMA-1007 avid NSBLs, to determine whether patients with NSBLs represent a higher risk clinical cohort, and to determine whether SUVmax can be used as a classifier of bone metastasis. Methods A retrospective audit of 214 men with prostate cancer was performed to investigate the clinical outcomes of [18F] PSMA-1007 avid NSBLs according to defined criteria. We also compared the serum PSA, Gleason score, and uptake time of patients with [18F]PSMA-1007 avid NSBLs to patients without [18F]PSMA-1007 avid bone lesions. Finally, we analysed an SUVmax threshold to identify bone metastases using ROC curve analysis. Results Ninety-four of 214 patients (43.9%) demonstrated at least one NSBL. No [18F]PSMA-1007 avid NSBLs met criteria for a likely malignant or definitely malignant lesion after a median 15.8-month follow-up interval (11.9% definitely benign, 50.3% likely benign, and 37.7% equivocal). There were no statistically significant differences in serum PSA, Gleason score, and uptake time between patients with [18F]PSMA-1007 avid NSBLs and those without [18F]PSMA-1007 avid bone lesions. All NSBLs with adequate follow-up had SUVmax ≤ 11.1. The value of the highest SUVmax distinguished between NSBLs and definite prostate cancer bone metastases, whereby an SUVmax threshold of ≥ 7.2 maximized the Youden’s index. Conclusion [18F]PSMA-1007 avid NSBLs rarely represent prostate cancer bone metastases. When identified in the absence of definite metastatic disease elsewhere, it is appropriate to classify those with SUVmax < 7.2 as likely benign. NSBLs with SUVmax 7.2–11.1 may be classified as equivocal or metastatic, with patient clinical risk factors, scan appearance, and potential management implications used to guide interpretation.


Introduction
Prostate-specific membrane antigen (PSMA) is a transmembrane cell surface protein which is overexpressed on prostate cancer cells. [ 68 Ga]Ga-PSMA-11 PET/CT imaging has greater diagnostic accuracy, management impact, and fewer equivocal findings than conventional imaging with CT and bone scan in staging men with high risk prostate cancer [1]. [ 18 F]PSMA-1007 offers multiple potential advantages over [ 68 Ga]Ga-PSMA-11 [2]. Firstly, 18 F has a lower positron energy than 68 Ga, theoretically improving spatial resolution and target to background ratio. [ 18 F]PSMA-1007 is excreted primarily in bile rather than urine and may increase reporter confidence in interpreting small locoregional lesions adjacent to the urinary tract in the absence of intensely avid radiourine present with [ 68 Ga]Ga-PSMA-11 [3]. Although this advantage of [ 18 F]PSMA-1007 could be reduced by the well-described technique of using furosemide forced diuresis with [ 68 Ga] Ga-PSMA-11 PET/CT [4][5][6][7][8], no direct comparative study with [ 18 F]PSMA-1007 has been performed. Finally, the [ 18 F] radionuclide is cyclotron rather than generator produced and has a longer half-life (110 min versus 68 min for [ 68 Ga]Ga-PSMA-11), which more readily enables a scalable centralized manufacturing process with distribution to peripheral centres. Although a centralized radiopharmacy model for [ 68 Ga]Ga-PSMA-11 distribution may be possible from cyclotron-based production of [ 68 Ga]GaCl 3 , reports of the amount of radiopharmaceutical produced are currently similar to the product activity when multiple 68 Ge/ 68 Ga generators are eluted in serial to a labelling platform [9]. [ 18 F]PSMA-1007 has been reported to be non-inferior [10,11] and potentially superior [12][13][14] to [ 68 Ga]Ga-PSMA-11 for lesion detection sensitivity in patients with prostate cancer. Indeed, a recent network meta-analysis found a potentially higher detection rate for [ 18 [15]. One potential disadvantage is the reported increased incidence of [ 18

Retrospective patient population
Patients who underwent [ 18 F]PSMA-1007 PET/CT at the Royal Brisbane and Women's Hospital, a large tertiary referral hospital, between May 2018 to November 2019 were retrospectively considered (Fig. 1). Inclusion criteria were men with histologically proven prostate cancer. Exclusion criteria were patients with [ 18 F]PSMA-1007 avid bone metastases, defined as PSMA-RADS 4 or 5 bone lesions [17] which were unequivocally reported as bone metastases by two experienced specialists. Indications for [ 18 F] PSMA-1007 PET-CT were (i) histologically proven prostate carcinoma staging prior to radiation therapy or surgery in intermediate or high-risk disease (i.e. PSA > 20, or Gleason score ≥ 4 + 3 = 7, or clinically ≥ T3), or (ii) restaging in the context of biochemical recurrence (PSA ≥ 0.2 ng/ ml post prostatectomy or ≥ 2.0 ng/ml above the nadir post radiotherapy). The Gleason score at diagnosis, previous and/ or current treatment, and most recent PSA value were also recorded. This audit was approved by the institutional review board of the Royal Brisbane and Women's Hospital, and the need for written informed consent was waived.
[ 18 F]PSMA-1007 manufacture and quality control [ 18 F]PSMA-1007 was manufactured on site using a kitbased approach on GE FASTlab or MX Tracerlab platforms, with comprehensive quality control testing of each batch to monograph standards for other F-18 radiopharmaceuticals (pH, chemical and radiochemical purity, radionuclide identity and purity, residual solvents, sterilising filter integrity, endotoxin, and sterility). Product radiochemical purity was measured by high performance liquid chromatography (HPLC) with a pre-defined acceptance criterion of > 95% [ 18 F]PSMA-1007. In separate experiments, free fluoride was also injected in our HPLC system to determine retention time and limit of detection for our instrument. While small radiopeaks (cumulatively less 5%) were observed close to the retention time of the product, no radiopeaks were observed corresponding to the retention time of free fluoride, excluding the presence of this impurity. In site validation processes, TLC tests were undertaken in parallel and no free fluoride was detected, further supporting the quantitative results of the HPLC testing. A product expiry of seven hours was assigned based on testing from end of synthesis.  [10]), and all patients were scanned (median uptake time 126 min, range 119-137 min) on a Biograph mCT scanner (Siemens Medical Solutions) from vertex to thighs. A lowdose non-diagnostic CT was also performed for the purpose of anatomical localisation and attenuation correction. Iodinated contrast was not used. PET scans were acquired in three-dimensional mode with an acquisition time of 2.5 min per bed position and a 4-min bed position over the pelvis. Emission data were corrected for randoms, dead time, scatter, and attenuation and were reconstructed iteratively using ordered-subsets expectation maximization (3 iterations, 21 subsets) with time of flight and point-spread function resolution recovery, followed by a post-reconstruction smoothing Gaussian filter (5 mm in full width at half maximum).

Image analysis
All images were dual-reported by two experienced nuclear medicine specialists using Syngo Via software (Siemens Medical Solutions). The [ 18 F]PSMA-1007 tracer replaced [ 68 Ga]Ga-PSMA-11 at our institution for clinical use at the start of the study period. The presence or absence of [ 18 F]PSMA-1007 avid NSBLs, defined as PSMA-RADS 3B lesions (i.e. equivocal uptake in a bone lesion not definitive but also not atypical of prostate cancer on anatomical imaging, including uptake without an anatomical correlate on CT) [17], were identified on reports, and the scans were retrospectively reviewed. If present, the number (1, 2, 3, 4, 5, > 5), regional site(s) (skull, cervico-thoraco-lumbar spine, pelvis including sacrum, ribs, sternum, pectoral girdle, upper limbs, lower limbs), and SUVmax of these lesions were recorded. The SUVmax was determined by placing a volume of interest encompassing the target lesion.

Retrospective subgroups
Three subgroups were identified from the retrospective cohort for further analysis (  (Table 1). Definite bone metastases were acquired from this separate prospective study rather than from the larger retrospective cohort to avoid confounding by reporting bias. Although PSMA-RADS 5 bone metastases are reliably identifiable due to corresponding anatomical findings on CT (i.e. typical sclerotic/ blastic change), there is no validated SUVmax threshold to distinguish between equivocal PSMA-RADS 3B ('equivocal uptake') and metastatic PSMA-RADS 4 ('high uptake') bone lesions using [ 18 F]PSMA-1007 PET/CT. Comparison to a cohort of 'metastases' arbitrarily defined by the same reporting clinicians would have been biased and self-fulfilling. Therefore concordant suspicious bone avidity independently identified with both agents was used to identify definite bone metastases for the purpose of our comparative analysis, noting that equivocal bone lesions on [ 18 F]PSMA-1007 PET/CT are typically non-avid on [ 68 Ga]Ga-PSMA-11 PET/CT [3]. Imaging in the prospective study was performed according to the same protocol on the same PET camera as described above for the retrospective cohort. Apart from age ≥ 60 years in the prospective cohort, inclusion criteria were the same as the retrospective cohort for primary staging and biochemical recurrence indications, with an additional group imaged for restaging of metastatic disease. There were no overlapping patients between retrospective and prospective cohorts. The SUVmax of the most avid concordant bone metastasis on [ 18 F]PSMA-1007 PET/CT in each patient was evaluated and recorded by an experienced nuclear medicine specialist, for the purpose of comparison to the most avid NSBL in the retrospective group (highest SUVmax measurement per patient) and for classification of bone metastasis. STATA 15[19] was used for all analyses. Data were described as mean and standard deviation (SD) for continuous normally distributed variables, median and interquartile range (IQR) for continuous non-normally distributed variables, and frequency and percent for categorical variables. Associations between categorical variables were assessed with either chi-square tests or Fisher's exact tests when the percentage of cells with expected counts < 5 was > 20%. Continuous normally distributed variables were compared between groups with t-tests, and continuous non-normally distributed variables were compared between groups with Mann-Whitney U tests. To further examine the relationship between the presence or absence of NSBLs and serum PSA, Gleason score and uptake time, regression analyses were performed. Specifically, linear regression analyses of the log transformed outcomes of PSA and uptake time and a binary logistic regression of the outcome of total Gleason score (8-10 versus < 8) were undertaken where NSBL status was included solely in the models, then was adjusted for scan indication (i.e. primary staging or biochemical recurrence), a clinically relevant covariate. Statistical significance was indicated where a p-value was < 0.05. An ROC curve analysis was performed to assess the classification performance of SUVmax in distinguishing NSBLs in Group B from bone metastases in the Group D. A cut-point maximising the sensitivity and specificity was derived using Youden's index (sensitivity + specificity -1). This ROC curve analysis excluded patients from the retrospective group aged < 60 years (n = 6) to align with the inclusion criteria of the prospective trial cohort.    Fig. 1). There were no statistically significant differences in baseline characteristics between those patients with and without follow-up (Table 2). These 77 patients had a total of 159 NSBLs (median SUVmax 3.4, IQR 2.7-4.4, range 1.8-11.1), and according to a per lesion analysis, no NSBLs met the defined endpoint criteria for a likely or definitely malignant lesion during follow-up (see Fig. 2). Most lesions were likely benign (80/159, 50.3%) or definitely benign (19/159, 11.9%), while 60/159 remained equivocal (37.7%). Subsequent per patient analysis (based on the most investigated NSBL) was similar to the per lesion analysis (likely benign 41/77, 53.2%; definitely benign 10/77, 13%; equivocal 26/77, 33.8%).

Clinical factors according to NSBL status
The median serum PSA and total Gleason scores of patients with NSBLs (Group A, n = 94) were similar to patients without NSBLs (Group C, n = 120) (p = 0.91, p = 0.23 respectively; Table 3). The median uptake time of patients with and without NSBLs were also similar (p = 0.77; Table 3). Unadjusted regression analyses and regression analyses adjusted for scan indication (i.e. primary staging or biochemical recurrence) were similar, and there remained no associations between patient NSBL status (i.e. NSBL present or absent) and the outcomes of serum PSA, Gleason score and uptake time (Supplementary Table 1).  (Fig. 3). ROC curve analysis showed that the highest SUVmax recorded distinguished between NSBLs and definite prostate cancer bone metastases with an area under the curve of 99.5% (95% confidence interval 98.3-100%, Fig. 3). An SUVmax cut-point of ≥ 7.2 maximized the Youden's index (sensitivity 100% and specificity 98.6%). Only 5/159 NSBLs with follow-up (Group B) (3.1%) had an SUVmax ≥ 7.2, and of these 1/5 was definitely benign, 3/5 were likely benign, and 1/5 was equivocal.  We also found that serum PSA levels and Gleason scores were similar in consecutive patients with and without [ 18 F] PSMA-1007 avid NSBLs, further suggesting that patients with NSBLs do not represent a higher risk cohort. In our series, ribs were by far the most common NSBL location, representing 60.4% of all NSBLs, consistent with previous reports [16,20]. Rib metastases are known to occur in prostate cancer, but they only account for 12-20% of bone metastases in autopsy studies, with most metastases occurring in the spine [22,23]. Only around 10% of prostate cancer bone metastases are solitary [24], and rib metastases are thought to be extremely rare in the absence of vertebral metastases [25]. A recent retrospective cohort study of 62 men with a solitary avid rib lesion on staging [ 68 Ga]Ga-PSMA-11 PET found that the vast majority of these lesions (61/62, 98.4%) met criteria for a benign lesion, and the authors recommended against percutaneous rib biopsy in this scenario unless there are high risk clinical features [26].

Discussion
The aetiology of [ 18 F]PSMA-1007 avid NSBLs remains uncertain. The possibility of free 18 F accounting for NSBLs in this study is highly unlikely due to stringent quality control measures (described in methods), similar uptake times between patients with NSBLs and those without any avid bone lesions, and the lack of diffuse skeletal uptake. Alternative malignant marrow pathologies are also unlikely because the ribs have minimal marrow production compared to other skeletal sites but showed a disproportionately high proportion of NSBLs. Marrow-based pathologies like myeloma, myelodysplastic syndrome, and other myeloproliferative neoplasms would also often result in diffuse rather than focal marrow involvement and would not be expected in over 40% of men. A benign cause of NSBLs seems more likely, and benign avid bone lesions are well described, with PSMA uptake having been demonstrated in healing fractures, osteoblastic activity, Paget's disease, haemangiomas, and fibrous dysplasia using [ 68 Ga]Ga-PSMA-11 or [ 18 F] DCFPyL [27][28][29][30]. These lesions are usually distinguished from typical sclerotic prostate cancer bone metastases on anatomical imaging and pose few problems for experienced readers. However, the presence of mild to moderate focal bone avidity without a typical morphological correlate on anatomical imaging is challenging because it is difficult to exclude atypical or low-volume bone metastases.
We identified multiple cases in which [ 18 [3]. Small molecule PSMA inhibitors such as [ 68 Ga]Ga-PSMA-11 and [ 18 F]PSMA-1007 share the glutamate-urea-lysine pharmacophore, which targets the catalytic domain of PSMA. Structurally, they differ in how they are conjugated to linking groups and radiometal chelators (e.g. HBED-CC) or prosthetic groups. The much higher incidence of NSBLs on [ 18 F] PSMA-1007 PET/CT with comparison to the other PSMA tracers in routine clinical use suggests an issue that is inherent to the molecule itself, possibly related to its known high binding affinity and internalization ratio [31].
With comparison to a group of definite [ 18 F]PSMA-1007 avid prostate cancer bone metastases (Group D, defined by concordant focal uptake independently identified on [ 68 Ga] Fig. 3 A SUVmax of most avid NSBL from Group B compared to definite bone metastases from Group D (each point represents a measurement for the highest bone lesion SUVmax within a single patient). B ROC curve of highest SUVmax within a patient as a classifier of NSBL from Group B compared to definite bone metastases from Group D Ga-PSMA-11 PET), the highest SUVmax of a skeletal lesion recorded within a patient was able to distinguish between NSBLs and bone metastases. If formulating a clinical report for an [ 18 F]PSMA-1007 PET/CT study, we therefore suggest that the SUVmax of the 'hottest' [ 18 F]PSMA-1007 avid NSBL (i.e. lesion without characteristic sclerotic/blastic changes) can measured to determine the likelihood of bone metastasis in patients who otherwise have no distant metastatic disease. Based on our findings we suggest classifying NSBLs with SUVmax < 7.2 as likely benign. The 5 NSBLs in our cohort with follow-up and SUVmax 7.2-11.1 are also most likely benign according to our findings; however there are insufficient numbers to provide greater certainty within this SUVmax range, and given that metastatic lesions may have a similar SUVmax, we recommend classifying these as either equivocal or metastatic, depending curative surgical treatment only, and post-operative serum PSA was undetectable. This NSBL was therefore classified as definitely benign on endpoint analysis upon the context of scan appearance (noting that NSBLs are frequently found in the ribs and isolated rib prostate cancer metastases are rare), clinical risk factors (including PSA level, ISUP grade, known extra-prostatic disease), and potential management implications to guide interpretation. For example, an equivocal lesion on an initial staging scan may not preclude an attempt at curative therapy; however a lower threshold for intervention may apply when restaging for biochemical recurrence. It should also be noted that while the PSMA-RADS classification system does take into account the number of lesions when assigning a classification about the level of suspicion, the PROMISE miTNM reporting framework uses lower PSMA uptake thresholds for classifying non-specific bone uptake as suspicious when The patient proceeded to local curative surgical treatment only, and post-operative serum PSA was undetectable. These NSBLs were therefore classified as definitely benign on endpoint analysis there are multiple foci versus a single focus [32]. Although the PROMISE system has not been validated for [ 18 F] PSMA-1007, the presence of multiple NSBLs may increase suspicion of metastatic disease in equivocal scenarios.
Clinical context is important to the interpretation of NSBLs, and all NSBLs in our study were clinically managed as benign. Clearly the presence of a NSBL is irrelevant to patient management in the context of widespread metastatic disease, and no further investigation is warranted. However, the interpretation of a NSBL as a metastatic site in a patient with otherwise only locoregional disease would reclassify the patient as having oligometastatic disease and potentially preclude curative intent treatment. It has recently been shown that local prostate radiotherapy with planned systemic therapy in patients with prostate cancer oligometastases (≤ 3 sites) improves overall survival [33], and notably > 85% of patients in our study had ≤ 3 NSBLs. Therefore, clinicians may consider proceeding with planned prostate radiotherapy in conjunction with androgen deprivation therapy in the context of equivocal [ 18 F]PSMA-1007 avid skeletal findings and repeat the PET in 3-6 months to assess the progress of the baseline NSBL. For high-risk patients in the biochemical recurrence setting, re-evaluation with PSMA PET/CT in 6 months may be appropriate. True metastatic disease is likely to have either responded or progressed in response to therapy [1].
The current study was limited by its retrospective nature and the small number of NSBLs with a histopathological gold standard. The inclusion of scans performed for both primary staging and biochemical recurrence of prostate cancer may represent an overly heterogeneous population, but it was considered legitimate given the research question investigating a potentially incidental finding. To examine the extent to which this influenced the results, analyses were performed with and without adjusting for scan indication, and the results did not change. The SUVmax statistical cut-point analysis was also limited by the small sample size of the prospective comparison group of definite bone metastases (n = 13). It is also important to note that the majority of the definite bone metastases (11/13) used in the comparative analysis had typical sclerotic/blastic changes on CT. Because PSMA PET is known to be more sensitive than CT in detecting early or low volume bone marrow metastases before the secondary effects of cortical or trabecular change are apparent [1,34], the important issue of determining a cut-point for PET positive but CT negative bone marrow metastases (which would presumably also be evident on MRI) requires further evaluation. We also acknowledge that the clinical utility of single centre SUV cut-offs can be significantly limited and relies upon like comparison to our specific imaging protocol. It is well documented that inter-scanner variability (including within and between scanner models, especially comparing analogue to digital silicon photomultiplier-based systems), as well as differences in image acquisition and reconstruction protocols can produce significant variability in SUV values [35][36][37]. Finally, it is possible that a longer follow-up period could see some NSBLs re-classified as true prostate cancer bone metastases. Indeed, a portion of equivocal bone lesions have been shown to be true bone metastases using [ 68 Ga] Ga-PSMA-11 [20] and [ 18 F]DCFPyL [21], and there was some overlap between NSBLs and bone metastases in our equivocal SUVmax range of 7.2-11.1, meaning that the potential for misdiagnosis and subsequent over-or undertreatment does exist in this range. Ultimately the comparatively high incidence of NSBLs on [ 18 F]PSMA-1007 PET/ CT remains a disadvantage of the tracer, and while our findings provide some guidance for interpretation, further studies aiming to differentiate benign and malignant bone uptake are required.

Conclusion
Non-specific mild to moderate focal [ 18 F]PSMA-1007 bone uptake without a typical morphological correlate on CT is common, and our findings suggest that these foci rarely represent prostate cancer metastases. When identified in the absence of definite bone metastases elsewhere, our results indicate that it is appropriate to classify those with SUVmax < 7.2 as likely benign. NSBLs with SUVmax 7.2-11.1 may be classified as equivocal or metastatic, with patient clinical risk factors, scan appearance, and potential management implications used to guide interpretation in this situation.