Optimization of PET protocol and interrater reliability of ¹⁸F-PSMA-11 imaging of prostate cancer

In recent years, prostate-specific membrane antigen (PSMA) has been the most widely studied target for imaging of recurrent and metastatic prostate cancer. ¹⁸F-PSMA-11, a fluorine-18 derivative of the frequently used ⁶⁸Ga-PSMA-11 PET radiotracer, was developed [1, 2], automatized [3] and evaluated for safety, biodistribution and dosimetry in a previously published study [4]. Before comparing the clinical efficacy of ¹⁸F-PSMA-11 to other PSMA tracers, a Phase 2 trial should be conducted to determine the scan protocol which will be applied in following studies and clinical practice. Phase 2 trial designs are usually exploratory where multiple scan parameters are tested to determine the optimal scan protocol. These parameters include radiotracer dosage, start of PET acquisition post injection, scan duration, image reconstruction parameters and updating the safety database. This requires the selection of the appropriate patient population. Finally, the variability between observers should be evaluated as the correct interpretation of PET images is a crucial step in the validation of the clinical efficacy of ¹⁸F-PSMA-11 and should lead to the development of criteria for image evaluation. The optimized scan protocol can then be applied in Phase 3 clinical trials to obtain data necessary for approval for application of the radiotracer in clinical use. In these larger studies, it is possible to evaluate the effect of the specific activity and the radionuclide on image quality, as well as the amount of nonradioactive ligand and carrier [5, 6]. The aim of this study was to determine the optimal scan protocol with regard to radiotracer dosage, acquisition time and duration and co-administration of furosemide. As a secondary objective, the interrater reliability was assessed.

All procedures performed in this study involving human participants were in accordance with the ethical standards of the Ethics Committee of the Ghent University Hospital (2017/1294) and with the 1964 Helsinki Declaration and its later amendments or comparable ethical standards (EudraCT nr, 2017-003461-96). The study was supported by the Flemish Foundation FWO TBM (T001517).

An increased interest in PSMA-based radiotracers has led to multiple clinical trials on PSMA PET imaging for prostate cancer [11‐13]. To achieve an accurate and reliable image interpretation, a standardized scan protocol is necessary for implementation in clinical practice [14]. This study determines the optimal scan protocol for PET imaging of ¹⁸F-PSMA-11 by considering the effect of activity dosage, acquisition time, scan duration and furosemide co-administration on image quality and lesion detection.

Various dosage schemes have been proposed for PSMA radiotracers. For ⁶⁸Ga-PSMA-11, the joint guidelines of EANM and SNMMI recommend a dosage of 2.0 ± 0.2 MBq/kg [14], while for ¹⁸F-PSMA-1007, dosages up to 4 MBq/kg are applied [15]. In this study, two dosage groups (2 and 4 MBq/kg body weight) were compared. The administered activity is the primary factor that determines the radiation exposure and should therefore be kept as low as reasonably achievable (ALARA principle) while minimizing the loss of diagnostic information [16‐18]. The higher dosage of 4 MBq/kg and associated increased radiation exposure will therefore only be tolerated when the difference in median score between both dosage groups is more than 1 point, as smaller differences will not be considered clinically relevant. Although the median difference between the two dosage groups was statistically significant, the > 1 point difference criterion was not achieved to accept the 4 MBq/kg body weight as dosage. Therefore, 2.0 ± 0.2 MBq/kg body weight is recommended for achieving images of sufficient quality while limiting the radiation exposure. This is in agreement with the joint EANM and SNMMI guidelines for PET/CT imaging with ⁶⁸Ga-PSMA [14]. For dosage optimization, inter-subject comparison was not possible as this would imply administration of both doses to each patient, which would unnecessarily increase patient radiation exposure. A compromise for this limitation is the comparison of the lesion detection of patients injected with 4 MBq/kg between the short (1.5 min/bp) and long (3 min/bp) scan time per bed position. One can argue that assuming linearity of the PET camera, imaging at half the acquisition time can simulate the administration of half the dosage. As seen in Table 5, the p value of the comparison between T60, 1.5 and 3 min and T180, 1.5 and 3 min is both times equal to 1, which confirms the recommendation of a dosage of 2 MBq/kg ¹⁸F-PSMA-11.

Many studies have compared early and delayed scanning; however, opposite results were obtained. Schmuck et al. and Afshar-Oromieh et al. both found a higher tumour-to-background ratio for some lesions on delayed ⁶⁸Ga-PSMA-11 imaging but did not agree on the change in overall detection rate [19, 20]. Rahbar et al. found no additional benefit in delayed ⁶⁸Ga-PSMA-11 acquisition [21], while Derlin et al. showed an improved image quality for ⁶⁸Ga-PSMA-I&T on delayed imaging but only when combined with delayed furosemide administration [22]. This shows that the acquisition time can be an important scan parameter but no consensus has yet been reached on this topic. As the longer half-life of ¹⁸F makes delayed imaging more accessible, it was deemed appropriate to evaluate imaging 1 h and 3 h after ¹⁸F-PSMA-11 administration. Images at T180 were more difficult to read and there appears to be increased uptake in the bone. However, a decrease in CNR was observed over time. This suggests that either the uptake in surrounding bone tissue increases and/or lesion uptake decreases. Additional analysis of the Phase 1 study [4] showed that the activity in bone versus total body increased with 10.86% from 90 to 300 min p.i. This increase is probably partially caused by free 18F. However, 90 min p.i., less than 10% of the injected activity is present in the blood of which only 22.2 ± 1.5% is free ¹⁸F-fluoride. This means that the amount of ¹⁸F-fluoride in absolute quantity is low. Another contributing factor could be slow tracer clearance from bone tissue, but this should be further investigated. This observation corresponds to a study by Afshar-Oromieh et al. who evaluated additional late scanning with ⁶⁸Ga-PSMA-11 [20]. Images at T60, 3 min/bp were given the highest median image quality scores and were therefore chosen as reference for following analyses. For the number of detected suspicious lesions, only T180, 1.5 min/bp was significantly inferior. Since only 42 lesions were detected in 11 patients, the number of lesions might be too small to detect a statistical difference between the other scan images. A high percentage agreement was only found between images at T60 (90.5%) while fewer lesions were detected on images at T180. Although the total number of detected suspicious lesions on T180, 3 min/bp were initially similar to the reference scan (32 vs 34 lesions respectively), retrospective analysis of the detected suspicious lesions showed that the suspicion of certain bone lesions was due to low image quality or aspecific uptake because of degenerate bone disease, which reduced the total number from 32 to 28. Finally, the relative CNR of suspicious lesions increased for longer acquisition times while decreasing when scanning occurred at a later time point. Taking into account all the abovementioned results, imaging 60 min p.i. with a scan duration of 3 min per bed position was recommended, which is in accordance to the joint EANM and SNMMI guidelines on imaging with ⁶⁸Ga-PSMA [14]. However, despite the lower image quality, the similar lesion detection rate of T60, 3 min/bp and T60, 1.5 min/bp would make the latter a possible acquisition scheme when shorter acquisition times are required. These results confirm similar previous findings by Goethals et al. [23] and Hausmann et al. [24]. No evidence was found to recommend delayed imaging with ¹⁸F-PSMA-11 in our study. No dynamic whole-body PET scans were performed for several reasons. Firstly, the previously conducted Phase 1 study [4] showed no increased uptake in major organs (except the kidneys) or possible lesions between 50 and 90 min p.i. Also, a broad set of inclusion criteria was applied to obtain a diverse cohort of patients in order to evaluate the radiotracer for different disease stages and metastases locations. Therefore, the condition of many elderly patients did not allow for intensive dynamic imaging protocols. Furthermore, the administration of a diuretic prevented prolonged scan duration. Therefore, the study protocol was limited to two acquisition times based on previously reported results [4] and literature reporting on scanning 60 min p.i. and increased lesion detectability 180 min p.i [14, 25, 26]. Dynamic imaging will be further investigated in ongoing animal experiments.

Furosemide co-administration reduced slightly the radiotracer interference on the ureters. However, the clinically relevant median difference of > 1 point was not achieved. This means that a good image quality can be obtained without the need for a diuretic which can cause additional discomfort. Nonetheless, the addition of furosemide can be beneficial for the detection of lymph nodes in the prostate region or in proximity of the ureters [22]. In this study protocol, a rather high dose of furosemide (40 mg) was administered 30 min before the start of the first PET/CT. It may be more appropriate to administer a lower furosemide dose simultaneously with the radiotracer administration for the comfort of the patient, which is also recommended in the guidelines of EANM and SNMMI for administration of ⁶⁸Ga-PSMA [14].

PSMA expression is not specific to the prostate only; physiological background uptake is seen in the kidneys, bladder, intestines, and salivary glands as well as in pathological conditions such as Paget’s bone disease and neovasculature of several solid tumours such as high-grade sarcomas [27‐31]. Moreover, some prostatic malignant lesions exhibit minimal PSMA expression and can be difficult to detect on PET [32]. Therefore, the interrater reliability can be a useful parameter to assess the clarity of ¹⁸F-PSMA-11 image interpretation and the potential need for gaining experience in differentiating prostate cancer lesions from benign focal uptake spots. The interrater reliability was assessed twice. Initially, the grade of agreement was determined in the original Phase 2 study within the 2 MBq/kg group on T60, 3 min/bp images. To see if a learning curve is applicable on the interpretation of ¹⁸F-PSMA-11 images, the analysis was repeated on images of the additional Phase 2 study which were carried out after all images from the initially included patients were evaluated. It was considered valuable to evaluate the ability to assign the correct disease status (local, locoregional, oligo- and polymetastatic disease) to each patient as this plays an important role in determining the appropriate treatment. The κ statistic of 0.92 (95% CI, 0.76–1) suggested an almost perfect interrater agreement in disease status and even increased to 1 (95% CI, 1–1) for the additional Phase 2. This would make ¹⁸F-PSMA-11 an important tool for knowing the disease status of the patient, which could influence the treatment management plan. The forest plot in Fig. 6 clearly shows improvement in each category where κ statistics were determined. This was in concordance with the interobserver agreement study by Fendler et al. [27] who also found that observers with less than 30 previous PSMA image readings only showed moderate interrater reliability contrary to more experienced observers who achieved substantial to almost perfect agreement. Retrospectively, all observed lesions from the original Phase 2 were discussed with both observers. Of the seven bone lesions on which the observers did not agree, two bone lesions were determined suspicious and subjected to follow-up while the other five bone lesions would not be made suspicious again after the gained experience from the trial.

This study was accompanied by some limitations. Firstly, determining the optimal dosage was solely based on a subjective scoring system. Furthermore, the PET characteristics also play a role in scan parameters. This was not accounted for as all scans were performed on one PET system (GE Healthcare). Finally, further optimization of image reconstruction parameters such as number of subsets, number of iterations and post-smoothing kernel should be investigated.

The best image quality was obtained by administration of 4.0 ± 0.4 MBq/kg ¹⁸F-PSMA-11. However, preference will be given to a dosage of 2.0 ± 0.2 MBq/kg due to the small difference in absolute score (max 1 point) and the ALARA principle. PET acquisition should start 1 h post injection with 3 min per bed position acquisition time. Depending on the medical history of the patient and the location of metastatic lesions, the co-administration of a diuretic can be useful. The increase of the κ value from 0.78 to 0.94 suggests that the interpretation of ¹⁸F-PSMA-11 is more reliable after 40+ PSMA scan readings.