Introduction

Prostate-specific membrane antigen (PSMA) positron emission tomography/computed tomography (PET/CT) has revolutionized the way prostate cancer is diagnosed and managed. The high level of PSMA expression in prostate cancer cells allows PSMA PET/CT to outperform conventional imaging in detecting not only recurrent and metastatic disease in the setting of biochemical failure, but also for initial staging of newly diagnosed intermediate- and high-risk prostate cancer [1]. Over the past decade, multiple clinical trials have demonstrated high accuracy of PET/CT using various PSMA-targeted radioligands and studies have shown that PSMA PET/CT findings impact the clinical management in approximately half of prostate cancer patients evaluated with such studies [2,3,4,5,6,7]. As a result, there has been widespread adoption of PSMA PET/CT across the globe along with integration into clinical practice guidelines [8, 9].

Despite its initial success, there is increasing awareness of several pitfalls of PSMA PET/CT, one of which is the presence of equivocal bone lesions with PSMA uptake. These often lead to further workup, and it has been suggested that many are ultimately benign (or represent a “false positive” finding for metastasis) [10]. There is also concern that certain PSMA-targeted radioligands such as [18F]F-PSMA-1007 – which was initially considered advantageous for better assessing the prostate (or prostatectomy bed) due to its predominant excretion through the liver, are associated with a higher frequency of indeterminate bone lesions due to “unspecific bone uptake” [11]. Ambiguity of the significance of these equivocal lesions interpreted as being malignant may in some cases exclude such patients from receiving curative-intent local definitive therapy (e.g., prostatectomy or radiation treatment). As a means to address this issue, standardized reporting schemes such as PSMA Reporting and Data System (PSMA-RADS), Prostate Cancer Molecular Imaging Standardized Evaluation (PROMISE), and European Association of Nuclear Medicine PSMA (E-PSMA) standardized reporting guidelines have been proposed to improve the interpretation and communication of findings on PSMA PET [12,13,14,15]. In PSMA-RADS, a bone lesion is assigned a score of “3B” if it demonstrates equivocal uptake on PET along with CT appearance which is not definitive but also not atypical for malignancy [12]. These lesions are recommended to undergo additional imaging, biopsy, or follow-up to determine their clinical significance. Nevertheless, the clinical significance of equivocal bone lesions on PSMA PET/CT and how to best use this information to manage the patient’s clinical treatment is still unclear.

Therefore, we conducted a systematic review and meta-analysis to evaluate (1) the prevalence of equivocal bone lesions on PSMA PET/CT, (2) the malignancy rate of equivocal bone lesions, and (3) the relationship between type of radioligand and equivocal bone lesions.

Methods

This systematic review and meta-analysis was conducted in accordance with the Preferred Reporting Items for Systematic Reviews and Meta-Analyses guidelines [16]. A study protocol was registered a priori to the Prospective Register of Systematic reviews (no. CRD42023486697) [17].

Search strategy

PubMed and EMBASE databases were searched up to December 11, 2023 to identify studies investigating equivocal bone lesions on PSMA PET/CT. The search query constructed based on key words (“prostate”, “PSMA”, “PET”, “equivocal”, and “bone”) and their synonyms was as follows: Prostate AND ((PSMA OR prostate-specific membrane antigen) AND (“positron emission tomography” OR PET)) AND (“PSMA-RADS” OR ((unspecific OR nonspecific OR equivocal OR indeterminate) AND (bone OR skeletal OR osseous))). The study selection process was performed by two reviewers (S.W. and H.A.V.) in consensus.

Inclusion and exclusion criteria

Studies were included if they were relevant to the Patient, Intervention, Comparator, Outcome framework [18]: (1) “patients” with prostate cancer (regardless of clinical setting), (2) PSMA PET/CT as “intervention” (regardless of type of PSMA-targeted radioligand) and interpretation with or without using PSMA-RADS, (3) No “comparator”, and (4) proportion of patients with equivocal bone lesions and their malignancy rates as “outcome”. We excluded studies if they met any of the following criteria: (1) publication types other than original articles or conference abstracts (e.g., review articles, editorials, and case reports), (2) cohort of < 10 patients, (3) insufficient information in the study to extract proportion relevant to the research question (i.e., prevalence of equivocal bone lesions and their malignancy rates), and (4) overlapping cohorts. When two or more studies were based on overlapping (or identical) cohorts, the one with more comprehensive and/or updated data was selected.

Data extraction and quality assessment

The following characteristics of the studies were tabulated using a standardized form: (1) study – author name(s), publication year, patient enrollment period, institution, country, design (prospective vs. retrospective); (2) clinical – age, clinical setting (e.g., primary staging, biochemical recurrence), Gleason score, prostate-specific antigen (PSA) level, reference standard for determining if equivocal bone lesions were malignant or not; and (3) PET – vendor, model, type of radioligand, injected dose, uptake time, acquisition time, definition of equivocal bone lesion (e.g., based on PSMA-RADS vs. other in-house criteria), number of patients with equivocal bone lesions, number of equivocal bone lesions, and number of equivocal bone lesions that were deemed malignant.

Assessments of the quality of the studies and their risk of bias were done based on Quality Assessment of Diagnostic Accuracy Studies-2 (QUADAS-2) [19]. All 4 domains (patient selection, index test, reference standard, and flow and timing) were assessed for studies evaluating the malignancy rates. However, only two domains (patient selection and index test) were assessed for studies that evaluated the prevalence of equivocal bone lesions. As evaluating the prevalence of equivocal bone lesions does not directly involve correlating them to a “reference standard”, the other two domains (reference standard and flow and timing) were not relevant for assessing quality and bias specifically for these studies. Both the data extraction and quality assessment were first done independently by two authors (S.W. and D.F.) and discrepancies were resolved by consensus with a third author (H.A.V.).

Data synthesis and analysis

The primary outcomes of this meta-analysis were: (1) the prevalence of equivocal bone lesions defined as the proportion of patients that had one or more equivocal bone lesions among all patients that underwent PSMA PET/CT and (2) the malignancy rate of equivocal bone lesions defined as the proportion of lesions that were deemed to be metastasis based on the reference standard among all evaluated equivocal bone lesions. The secondary outcomes were to explore heterogeneity by performing subgroup analyses, most importantly stratified to the type of PSMA-targeted radioligand ([18 F]PSMA-1007 vs. others [[68Ga]Ga-PSMA-11 or [18 F]DCFPyL], [18 F]PSMA-1007 vs. [68Ga]Ga-PSMA-11, and [18 F]PSMA-1007 vs. [18 F]DCFPyL]), but also by additionally relevant subgroups such as clinical setting (e.g., primary staging, biochemical recurrence, etc.) or definition of equivocal bone lesions (e.g., based on PSMA-RADS).

Pooling of the proportions and calculation of their 95% confidence intervals (CI) were done using a random-effects model with the “meta” package in statistical software R (version 4.2.2; R Foundation for Statistical Computing, Vienna, Austria) [20]. Heterogeneity was assessed with the Higgins I2 test [21]. Heterogeneity was explored by comparing pooled proportions of the subgroups [22]. Publication bias was assessed with the Egger test and visualized using the funnel plot [23]. P values < 0.05 were considered statistically significant.

Results

Literature search

Three-hundred and twenty-two articles were initially identified from the systematic database search. After removal of 96 duplicates and exclusion of 168 papers by screening of titles and abstracts, 44 articles were considered potentially eligible. Upon full-text review, 19 of these studies were excluded for the following reasons (Supplementary Table 1): study does not deal with equivocal bone lesions (n = 3); equivocal bone lesions and equivocal soft tissue lesions were not separately documented (n = 6); overlapping patient cohort (n = 8); and topic deals with equivocal bone lesions on PSMA PET/CT but details of neither their prevalence nor malignancy rates were provided (n = 2). Ultimately, 25 studies (4484 patients) were included in the systematic review and meta-analysis (24–48). Flowchart for study selection is shown in Fig. 1.

Fig. 1
figure 1

PRISMA flowchart summarizing study selection process. PRISMA = Preferred Reporting Items for Systematic Reviews and Meta-Analyses

Full size image

Characteristics of included studies

The characteristics of the included studies are summarized in Tables 1, 2 and 3. Among all 25 studies, 8 (32%) evaluated the prevalence of equivocal bone lesions only [28, 30,31,32, 35, 40, 42, 48], 3 (12%) the malignancy rate of equivocal bone lesions only [34, 45, 47], and the remaining 14 (56%) both the prevalence and the malignancy rates of equivocal bone lesions [24,25,26,27, 29, 33, 36,37,38,39, 41, 43, 44, 46]. Most studies were retrospective in design (n = 22, 88%) [24,25,26, 28,29,30,31,32,33,34,35, 37,38,39, 41,42,43,44,45,46,47,48] and performed at single centers (n = 23, 92%) [24,25,26,27,28, 30,31,32,33, 35,36,37,38,39,40,41,42,43,44,45,46,47,48]. PSMA PET was performed for primary staging in 6 studies (24%) [27, 33, 35, 37, 43, 48], biochemical recurrence in all or most of the patients in 6 (24%) [25, 26, 30, 34, 41, 45], and a mixed cohort in 9 (36%) [24, 28, 29, 32, 38,39,40, 46, 47]. In the remaining, one study (4%) specifically included patients with non-detectable PSA levels after RP and the remaining 3 (12%) did not report cohort characteristics [36]. [18F]PSMA-1007 was used in 13 studies (52%) [24,25,26,27,28,29,30, 32, 35,36,37, 44, 45], [68Ga]Ga-PSMA-11 in 6 (24%) [31, 34, 39, 42, 43, 48], [18F]DCFPyL in 3 (12%) [35, 38, 47], and a combination of [18F]PSMA-1007 and non-[18F]PSMA-1007 radioligands in 3 (12%) [33, 41, 46]. In 9 studies (36%), PSMA PET/CT was interpreted and equivocal bone lesions were accordingly defined per PSMA-RADS version 1.0 [24, 26, 33,34,35, 38, 43, 46, 47]. In the 16 remaining studies, various definitions were used (or details were not provided), most common being PSMA uptake without a morphologic correlate on CT (n = 5, 20%) [27, 32, 36, 37, 41]. Most of the studies that evaluated the malignancy rate of equivocal bone lesions (13/17 [76.5%]) used the best value comparator (BVC; a combination of imaging, clinical, histopathologic, or biochemical evaluations such as response to therapy) as the reference standard for determining whether the equivocal bone lesion was malignant or not [24, 25, 27, 29, 33, 34, 37,38,39, 41, 43, 46, 47]. The other studies used the following reference standards: biopsy (n = 1, 5.9%) [45], magnetic resonance imaging (MRI) (n = 2, 11.8%) [26, 44], and PSA kinetics (n = 1, 5.9%) [36].

Table 1 Study and patient characteristics
Full size table
Table 2 PSMA PET/CT characteristics
Full size table
Table 3 Subgroup analyses on the prevalence of equivocal bone lesions and their malignancy rates stratified to type of PSMA-targeted radioligand, methods for reporting PSMA PET/CT, and clinical setting
Full size table

Quality and risk of bias

The overall quality of the included studies was considered moderate to good, with 13 (76.5%) of the 17 studies evaluating malignancy rate of equivocal bone lesions satisfying 4 or more of the 7 QUADAS-2 domains and 5 (62.5%) of the 8 studies only evaluating the prevalence of equivocal bone lesions satisfying 3 or more of the 4 evaluated QUADAS-2 domains. The detailed breakdown of QUADAS-2 assessments are provided in Fig. 2.

Fig. 2
figure 2

Grouped bar charts summarizing risk of bias (left) and concern for applicability (right) in 25 included studies according to the QUADAS-2 tool. Individual study results are shown at the bottom. For 8 studies that only reported the prevalence of equivocal bone lesions on PSMA PET/CT (but not their malignancy rates), a modified approach using only 2 of the 4 domains (patient selection and index test) were assessed as the other 2 domains (reference standard and flow and timing) were not relevant

Full size image

For patient selection, 2 studies were considered at high risk for bias and had high concern for applicability: in one study, PSMA PET/CT was performed in patients that had undetectable PSA after prostatectomy [36] and another study only evaluated patients who were referred for PSMA PET-guided biopsies [45]. Another study had an unclear risk of bias as it was not explicitly mentioned whether patient enrollment was consecutive or not [37].

Regarding the index test, in the aforementioned study where patients had undetectable PSA after prostatectomy, readers were not blinded and were aware of this unique clinical setting, resulting in a high risk of bias [36]. In another study, PSMA PET was interpreted alone – specifically without correlation with CT – which not only negatively influences the diagnostic accuracy of PSMA PET (high risk of bias), but also is not applicable to current day practice where they are interpreted simultaneously (high concern for applicability) [39]. Most other studies (n = 12) were at unclear risk of bias as it was not explicitly stated whether PSMA PET/CT was interpreted blinded to the outcome.

In the reference standard domain, only one study had low risk of bias, as it exclusively was based on histopathological assessment of PSMA-PET guided biopsy [45]. However, this was considered to have high concern for applicability, as it would not be feasible to biopsy all equivocal bone lesions in daily practice. Three studies had high risk of bias as they were based on only MRI correlation (n = 2) [26, 44] or only PSA kinetics (n = 1) [36]. All 13 other studies were considered at unclear risk as they were based on the BVC.

Finally, in the flow and timing domain, those 13 studies (at unclear risk in the reference standard domain) were considered to be at high risk of bias since individual patients in the study had different reference standard by definition (i.e., BVC). Two additional studies were at unclear risk of bias as the interval between PSMA PET/CT and the reference standard was not explicit [26, 44].

Prevalence of equivocal bone lesions

Twenty-two studies reported the prevalence of equivocal bone lesions on PSMA PET/CT. The prevalence in each study ranged from 0 to 78% and the pooled prevalence across all 22 studies was 20% (95% CI: 12–31%) (Fig. 3). Substantial heterogeneity was present (I2 = 95.8%). There was no significant publication bias according to the Egger’s test (p = 0.21) and funnel plot (Supplementary Fig. 1A). At subgroup analyses, type of PSMA-targeted radioligand was associated with heterogeneity (p < 0.01), but not the clinical setting (p = 0.32) or how equivocal bone lesions were defined (p = 0.42) (Table 4). Specifically, studies that used [18F]PSMA-1007 (n = 14) were associated with a greater prevalence compared to those that used other radioligands ([68Ga]Ga-PSMA-11 and [18F]DCFPyL, n = 10): 36% (95% CI 26–48%) vs. 8% (95% CI 4–14%), respectively (Fig. 4). Subgroup analysis comparing [18F]PSMA-1007 with each of the radioligands are shown in Supplementary Fig. 2.

Fig. 3
figure 3

Forest plot showing pooled prevalence of equivocal bone lesions in all 22 studies. Events = number of patients with equivocal bone lesions. Total = number of patients in the study

Full size image
Fig. 4
figure 4

Forest plots showing pooled prevalence of equivocal bone lesions stratified to subgroups based on type of PSMA-targeted radioligand. Studies that used [18F]PSMA-1007 (n = 14) had a greater pooled prevalence compared to those that used other radioligands ([68Ga]Ga-PSMA-11 and [18F]DCFPyL, n = 10, p < 0.01)

Full size image

Malignancy rate of equivocal bone lesions

Seventeen studies reported the malignancy rate of equivocal bone lesions on PSMA PET/CT. The prevalence in each study ranged from 0 to 77% and the pooled prevalence across all 17 studies was 14% (95% CI: 7–25%) (Fig. 5). Substantial heterogeneity was present (I2 = 90.1%). There was no significant publication bias according to the Egger’s test (p = 0.52) and funnel plot (Supplementary Fig. 1B). At subgroup analyses, type of PSMA-targeted radioligand was associated with heterogeneity (p = 0.01), but not the clinical setting (p = 0.32) or how equivocal bone lesions were defined (p = 0.60) (Table 4). Specifically, studies that used [18F]PSMA-1007 (n = 12) were associated with a lower malignancy rate compared to those that used other radioligands ([68Ga]Ga-PSMA-11 and [18F]DCFPyL, n = 8): 8% (95% CI 3–19%) vs. 29% (95% CI 17–44%), respectively (Fig. 6). Subgroup analysis comparing [18F]PSMA-1007 with each of the radioligands are shown in Supplementary Fig. 3.

Fig. 5
figure 5

Forest plot showing pooled malignancy rate of equivocal bone lesions in all 17 studies. Events = number of malignant bone lesions. Total = number of equivocal bone lesions

Full size image
Fig. 6
figure 6

Forest plots showing pooled malignancy rate of equivocal bone lesions stratified to subgroups based on type of PSMA-targeted radioligand. Studies that used [18F]PSMA-1007 (n = 12) had a lower pooled malignancy rate compared to those that used other radioligands ([68Ga]Ga-PSMA-11 and [18F]DCFPyL, n = 8, p = 0.01)

Full size image

Discussion

In this systematic review and meta-analysis, we investigated the prevalence of equivocal bone lesions on PSMA PET/CT in patients with prostate cancer and how often they are malignant. Overall, equivocal bone lesions were often encountered, approximately in 1 of 5 patients (pooled prevalence = 20%, 95% CI: 12–31%). However, the rate of malignancy was low with approximately 1 in 7 equivocal bone lesions deemed to be metastatic disease (pooled malignancy rate = 14%, 95% CI: 7–25%). The results of our study substantiate on a larger scale, the trends we have been observing in smaller individual reports. Although PSMA PET/CT has revolutionized how we manage patients with prostate cancer, our study highlights that we need to keep in mind this important pitfall of equivocal bone lesions. There needs to be increased awareness to all stakeholders involved, including expert and community nuclear medicine physicians and radiologists who interpret PSMA PET/CT, referring physicians, and patients and caregivers in order to further optimize clinical recommendation based on the most accurate or reliable information which is clinically available [49].

Type of radioligand was a significant factor of heterogeneity between the included studies. Specifically, the prevalence of equivocal bone lesions was higher and their malignancy rate was lower in studies using [18F]PSMA-1007 (pooled estimate = 36%, 95% CI 26–48%) and 8%, 95% CI 3–19%, respectively) compared with other radioligands (pooled estimate = 8%, 95% CI 4–14% and 29%, 95% CI 17–44%, respectively). The reason for this difference between type of PSMA-targeted radioligand is not yet clearly understood. It has been suggested that the lower positron energy (and greater spatial resolution) and the longer half-life (and superior signal-to-background ratio) of [18F] compared to [68Ga] may partly explain this [29]; however, [18F]DCFPyL also is synthesized with [18F] and therefore this explanation cannot be considered valid. Others have suggested that it may be related to the higher binding affinity to receptors and internalization rate of [18F]PSMA-1007 [50, 51]. Regardless, the difference between [18F]PSMA-1007 and other radioligands has important clinical implications. [18F]PSMA-1007 has been considered advantageous for delineating the primary tumors and local recurrence after radical treatment and for radiotherapy planning (e.g., dose painting) due to its lack of urinary excretion compared with other radioligands [52,53,54]. Therefore, these characteristics related to radioligand type should be taken into consideration when choosing which one to use, and how to interpret them (in addition to jurisdictional constraints and access issues such as [18F]PSMA-1007 not yet approved in the United States and whether there is a cyclotron on site). Moreover, additional training may be needed before using [18F]PSMA-1007 to avoid overcalling these equivocal bone lesions.

Other factors such as clinical setting or how equivocal bone lesions were defined were not a substantial source of heterogeneity (p = 0.32–0.60). As bone metastases can virtually occur in any clinical setting, across the broad categories of primary staging, biochemical recurrence, and even during systemic treatment for known metastatic disease, the risk stratification itself (e.g., PSA level, PSA kinetics, or Gleason grade group) may be more important in determining whether the equivocal bone lesion is metastatic or not. For example, while the overall pooled malignancy rate was 14% (95% CI: 7–25%), few of the studies at the extreme ends warrant mention to illustrate this point. For example, two studies in which PSMA PET/CT was performed in predominantly high-risk patients – Paone et al. [37] (high risk in 58.8% [47/80], median PSA 23.7 ng/ml) and Simsek et al. [43] (high risk in 81.2% [289/356], median PSA 16.4 ng/mL) had very high malignancy rates of 76.9% (70/91) and 70.0% (7/10), respectively, substantially higher than the overall pooled malignancy rate of 14%. These data support tailoring the workup of equivocal bone lesions on PSMA PET/CT depending on the pretest likelihood of bone metastasis. They can be considered metastasis and managed accordingly in patients at high risk for metastasis (or with co-existing additional sites of definite metastasis elsewhere); on the contrary, in a patient at low risk for metastasis (especially who is a candidate for locoregional definitive treatment), higher thresholds should be applied for calling equivocal bone lesions as metastases so that the patient will not be denied potentially curative intent treatment. If risk is truly indeterminate, additional imaging, biopsy, or follow-up to confirm their significance may still be required. Further investigation is needed to see whether quantitative analysis of PSMA PET (e.g., maximum standardized uptake value [SUVmax] such as cutoff of > 7.2 suggested by Arnfield et al. [24]), secondary PSMA uptake patterns (e.g., focality and changes in late-phase acquisition [55]), more in-depth morphological analysis of CT correlates, or using additional tools such as genomic classifiers (e.g., Decipher test) will help identify metastasis or not in such scenarios [56, 57].

Although the definition of equivocal bone lesions (or how PSMA PET/CT was interpreted) was not a substantial factor of heterogeneity, this issue deserves more in-depth discussion. Nine (36.0%) of the studies used PSMA-RADS version 1.0, while various other definitions were used (or details not provided) in the remaining studies. No studies up to now have evaluated the recently updated PSMA-RADS version 2.0 [13]; however, only minimal impact on the prevalence and malignancy rate of equivocal bone lesions is to be expected considering the content of the modifications. It is unclear whether the conduct of using PSMA-RADS (or any other standardized reporting system) will in itself improve assessment of bone lesion. However, lessons can be learned from the history of the debate between Prostate Imaging Reporting and Data System (PI-RADS) versus Likert scale for interpreting MRI. While diagnostic performance was not significantly different (at least in high-volume tertiary centers), PI-RADS eventually became the standard for prostate MRI interpretation because it provided clear structure, standardization, and definition allowing the potential for decreasing inter-reader variability [58, 59]. PSMA-RADS could be used as a platform for research or utilized as part of clinical trials across different institutions and settings to help establish the clinical significance and most optimal way to manage equivocal bone lesions on PSMA PET/CT. Such structured reporting may also improve inter-reader agreement for PSMA PET/CT interpretation [60].

There were a few limitations in this systematic review and meta-analysis. First, there was substantial heterogeneity between the included studies. Nevertheless, we were able to demonstrate that at least some of this heterogeneity arises from the differences in type of PSMA-targeted radioligand ([18F]PSMA-1007 vs. others [[68Ga]Ga-PSMA-11 or [18F]DCFPyL]). Secondly, most (13/17 [76.5%]) of the included studies used BVC as the reference standard for determining whether the equivocal bone lesion was malignant or not. Although we acknowledged that this is a potential source for unclear risk of bias in the QUADAS-2 assessments, it should be emphasized that BVC is not perfect. As a matter of fact, using BVC could lead to both false positive and negative conclusions. Nonetheless, biopsy (used as the sole reference standard in 1 study [45]) cannot be used for all patients nor would that be ethically justified. Furthermore, interpretation of bone biopsies can be challenging especially when lesions are sclerotic, which is often the case in the setting of patients with prostate cancer, with approximately 1 in 4 biopsies yielding nondiagnostic results [61]. BVC is commonly used in clinical practice to decide whether equivocal bone lesions are metastasis or not. Therefore, the pooled estimates derived from this meta-analysis probably reflect real world evidence. Finally, we were unable to directly analyze the relationship between the location of the equivocal bone lesion and malignancy rate. Yet, it is well known from extensive literature that the likelihood of metastasis is different between locations – not only between axial and non-axial bones but even within axial bones (e.g., ribs vs. vertebrae or pelvic bones) [62, 63]. Although further investigation will be needed in this specific context, we can speculate that equivocal bone lesions on PSMA PET/CT will less likely be metastasis in the ribs, especially when in isolation without additional sites of disease in the pelvis or vertebrae.

In conclusion, equivocal bone lesions are often encountered on PSMA PET/CT but exhibit a low malignancy rate. Compared to other radioligands, [18F]PSMA-1007 requires special attention as it is associated with a higher frequency and lower rate of metastasis.