The accuracy and intra- and interobserver variability of PSMA PET/CT for the local staging of primary prostate cancer

Prostate cancer (PCa) is the most common cancer in men of middle and older age [1]. Adequate local staging (T-stage) in primary PCa is important to guide treatment decisions. Such as surgical planning (including nerve-sparing options) in patients undergoing radical prostatectomy or in those who opt for external radiotherapy of the prostate. Curation is more likely achieved when disease is limited to the prostate gland (i.e., T-stage ≤ 2).

The presence of T3-disease (i.e. T3a in case of extracapsular extension, or T3b in case of seminal vesicle invasion [2]) may be assessed by a combination of serum prostate specific antigen (PSA), digital rectal examination, transrectal ultrasound, the outcome of prostate biopsies and by magnetic resonance imaging (MRI). Although MRI is the most accurate single staging modality with high specificity (82–88%), it’s overall accuracy to detect T3-disease is hampered by its lower sensitivity (51–57%) [3, 4]. Therefore, efforts may be made to improve accuracy of detecting T3-disease.

Prostate-specific membrane antigen (PSMA) positron emission tomography / computed tomography (PET/CT) is increasingly being used for initial staging of patients with intermediate and high-risk primary PCa as well as in patients with recurrent disease after treatment with curative intent [5‐7]. While PSMA-PET is an adequate modality for staging metastases, it may also be applied for the assessment of the local extent of primary PCa. Several studies on the assessment of local staging on PSMA PET showed promising results [8]. However, these studies were performed in single centre settings and had a limited number of patients, thus possibly influencing the statistical robustness. Also, in these studies almost exclusively ⁶⁸Ga-labelled PSMA tracers were used, which may question its generalizability to the ¹⁸F-labelled PSMA tracers.

Therefore, we assessed a large, multicentre cohort of patients with intermediate and high-risk PCa using both ⁶⁸Ga- and ¹⁸F-labelled PSMA tracers. The diagnostic accuracy of molecular imaging local tumour staging by PSMA PET/CT (miT-stage) was determined with the radical prostatectomy specimen (pT-stage) serving as the reference standard. Furthermore, as an important determinant of diagnostic test reliability, both intra-observer and interobserver variability of PSMA PET/CT for local tumour staging were assessed.

This retrospective cohort study was conducted by the Amsterdam UMC and the Netherlands Cancer Institute – Antoni van Leeuwenhoek (NCI-AVL) as tertiary referral centres within the Prostate Cancer Network the Netherlands. Approval of the institutional review board of both hospitals was obtained for this study, meanwhile waiving the need to receive informed consent (VUmc2019.586 and IRBd20-041).

The presence or absence of locally advanced disease (≥ T3-disease) in patients with newly diagnosed PCa is of importance for risk stratification, therapeutic decision making, and for the prediction of oncological outcome. Besides conventional clinical variables such as the determination of PSA levels, clinical tumour stage based on digital rectal examination and prostatic transrectal ultrasound, prostate MRI has been established as a pivotal and most accurate diagnostic instrument to determine local tumour stage [7]. Although MRI has uniformly shown excellent specificity for locally advanced disease, the sensitivity of MRI to predict locally advanced tumour stage (≥ T3-disease) was reported to be limited and heterogeneous among studies [4]. Therefore, there is still an unmet need to improve the prediction of final pathological tumour stage in patients newly diagnosed with PCa. Recently, PSMA PET/CT received increasing attention regarding its value in detecting the dominant cancer lesion in patients with newly diagnosed PCa and for its diagnostic performance in predicting local disease extent [20]. In the present multicentre study, reporting on 600 patients undergoing RARP for locally confined PCa and who were preoperatively staged by PSMA PET/CT imaging, the diagnostic performance of PSMA PET/CT for local tumour stage in the radical prostatectomy specimen was studied as the reference standard. The sensitivity, specificity, and diagnostic accuracy (AUC) of PSMA PET/CT were 58%, 59% and 0.59 for pT3a-stage, 30%, 97% and 0.64 for ≥ pT3b-stage, and 68%, 61% and 0.64 for overall ≥ pT3-stage disease.

PSMA PET/CT is increasingly used as a metastatic screening tool in patients with a primary diagnosis of PCa [7] and may indeed be of value for the assessment of the local extent of the primary prostate tumour as well. Woo et al. included in a systematic review and meta-analysis a total of twelve studies: ten prospective single centre studies, one prospective single centre study and one dual centre study on the PET assessment of local staging [8]. The reported pooled sensitivity and specificity of PSMA PET, using exclusively ⁶⁸Ga-based PSMA PET tracers, were 72% (95% CI 0.56–0.84) and 87% (95% CI 0.72–0.94) for pT3a-stage disease, and 69% (95% CI 053–0.81) and 94% (95% CI 0.90–0.96) for pT3b-stage disease, respectively. The number of patients in these studies ranged from 21 to 140 and therefore was relatively small. The authors further showed that PSMA PET/MRI might have an improved diagnosed performance for local tumour stage compared to PSMA PET/CT imaging, although the number of studies using PSMA PET/MRI was only low. It is reasonable that combining different imaging modalities improves the diagnostic performance over a single imaging modality. This meta-analysis does not, however, address which imaging modality contributed preferentially to the observed improved diagnostic performance for predicting local tumour stage as compared to either of the two single imaging modalities alone. A further setback of the included studies in the meta-analysis is that often PSMA PET/CT scans were assessed by a single observer only and observer agreement scores were not provided, as opposed to the present study. This is important as interpretation of scans might differ substantially between observers. Finally, the meta-analysis by Woo et al. [8] only addressed the diagnostic performance of ⁶⁸Ga-labelled PSMA radioligands and not that of ¹⁸Fluoride-labelled PSMA radioligands. This is a limitation as a large proportion of the presently performed PSMA PET scans are based on ¹⁸Fluoride-labelled radioligands, e.g., [¹⁸F]-JK-PSMA-7, [¹⁸F]DCFPyl, and [¹⁸F]PSMA-1007.

In our study, the diagnostic accuracy of PSMA PET/CT for the prediction of locally advanced disease (≥ pT3 stage) was lower than reported in smaller, initial studies and in the meta-analysis by Woo et al. [8], but was comparable to that in the recent, larger prospective study by Sonni et al. [21]. In this study comparing PSMA PET, MRI and histopathology for local tumour staging in 74 patients, the AUC for detection of extracapsular extension (ECE) was 0.59 and 0.63 for seminal vesical invasion (SVI).

Differences in diagnostic performances between studies may be due to differences in study population, differences in study set-up, differences in PSMA-radioligands and PET image acquisition, differences in the interpretation and reporting of PSMA PET/CT-based disease parameters, and in differences in the reporting of radical prostatectomy specimens. Possibly a more controlled single centre, single PSMA-radioligand, and single PET/CT system setting may have led to higher accuracy rates for predicting locally advanced disease as reported in initial PSMA PET/CT studies performed by Fendler et al. (accuracy 71% for ECE and 86% for SVI) and von Klot et al. (sensitivity and specificity 90% and 90% for ECE, respectively 75% and 100% for SVI [8, 17, 22]. It remains yet largely unknown which of the above listed variables are responsible for the reported differences in diagnostic performance between studies.

In the present study, the observers used a consensus-based standardized reporting system for molecular imaging local tumour stage, which optimized agreement of interpretation. However, relevant differences were found between intra-observer and interobserver agreement rates. The intra-observer agreement for PSMA PET/CT assessment of overall ≥ T3 stage was substantial (k 0.69), whereas the interobserver agreement for the assessment of overall ≥ T3 stage was moderate (k 0.45). Likewise, a substantial intra-observer agreement on ≥ T3b stage (k 0.75) was found, whereas the interobserver agreement for the assessment of ≥ T3b stage was moderate (k 0.41). The substantial intra-observer agreement scores suggest that further improvements in interobserver agreement may be reached by stricter adherence to image interpretation consensus criteria, which may add to the reliability of PSMA PET/CT for local tumour staging.

In comparison, Sonni et al. using a 5-point Likert scale found poor interobserver reliability rates for PSMA PET/CT assessment of ECE (intraclass correlation coefficient (ICC) 0.203) and SVI (ICC 0.081) [21]. In this study the T-staging was assessed visually in a binary manner, leaving no room for equivocal results which may have contributed to the lower interobserver agreement rates compared to our results. Muehlematter et al. reported a fair interobserver reliability for ECE (0.40; 95% CI 0.33–0.47) and SVI assessment (0.33 (95% CI 0.17–0.25) for PSMA PET/MRI, which did not differ significantly from MRI as single modality [23].

In theory, local prostate cancer staging may be hampered by the vicinity of intense urinary tracer activity in the bladder. Therefore, there may be an advantage in local prostate cancer staging for PSMA PET tracers that exhibit lower urinary excretion such as [¹⁸F]PSMA-1007, compared to the primarily urinary excreted tracers [¹⁸F]DCFPyL, [¹⁸F]-JK-PSMA-7 and [⁶⁸Ga]Ga-PSMA-11. We found a lower sensitivity and diagnostic accuracy of [¹⁸F]DCFPyL versus [⁶⁸Ga]Ga-PSMA-11 specifically for detecting pT3a disease. This possibly may be attributed to higher urinary tracer activity in [¹⁸F]DCFPyL compared to [⁶⁸Ga]Ga-PSMA-11 [24], but other factors may play a role as well. Moreover, the difference was not statistically significant, so no definitive conclusions can be drawn from these data.

Earlier studies point to an advantage of lower urinary activity levels in local staging performance mainly in the biochemical recurrence setting (especially after prostatectomy), and not so much in the primary setting with the prostate still in situ [25‐29]. In our experience, T-stage assessment in the primary setting is not hampered by urinary activity in most cases. Still, the occasional smaller tumour at the basal part of the prostate or tumours with bladder invasion may be better staged with lower urinary activity levels. Therefore, strategies to reduce urinary activity (including tracer selection and forced diuresis) in PSMA PET [24‐31] may be beneficial in miT-stage assessment as well.

Strengths of the current study are the number of included patients and the comparison with histopathology as reference standard. Furthermore, patients were includedin a high-volume, dedicated multicentre prostate cancer network with a prospectively maintained database and analysed by nuclear medicine physicians with ample experience in PSMA PET reading according to E-PSMA guidelines [14]. The mix of different tracers, protocols and PET/CT systems reflects variations among clinical practices and increases generalizability of our results. Next to the strength of this analysis, it is also one of the limitations of our study. As we have used real-world data with different PSMA tracers, PET/CT systems and protocols this may have influenced the outcomes. However, PET/CT systems, protocols and image reconstructions adhered to EARL standards and EANM guidelines [10‐12, 16], keeping data heterogeneity within boundaries. Other limitations include the retrospective character of this study, the observation that only a subset of 100 PSMA PET-scans were re-assessed and dually assessed which may have led to larger confidence intervals in the reported observer agreement rates, and the fact that a small proportion (3.5%) of patients were excluded because of a PSMA negative primary tumour, as it is yet unknown whether this apparent absence of PSMA expression is related to imaging technical factors or whether it is due to tumour-associated factors. Differences in the case-mix offered to the nuclear medicine physicians may have influenced the interobserver agreement rates.

Addition of PSMA PET/CT based local tumour staging might contribute to the existing nomograms for the prediction of locally advanced disease in the radical prostatectomy specimen (i.e., pT-stage disease). Further research may be aimed at improving qualitative interpretation criteria as well as possible addition of quantitative parameters which may improve diagnostic accuracy and interobserver agreement rates.

In a large, multicentre study evaluating 600 patients with newly diagnosed with intermediate and high-risk prostate cancer, we showed that PSMA PET/CT may have a valuable role in local tumour staging when pathological tumour stage in the radical prostatectomy specimen was used as the reference standard. The interobserver and intra-observer variability of tumour extent on PSMA PET/CT was moderate to substantial.