Using ⁶⁸Ga-PSMA-11 PET/CT for Therapy Response Assessment in Patients with Metastatic Castration-Resistant Prostate Cancer: Application of EAU/EANM Recommendations in Clinical Practice

DISCUSSION

Despite EAU/EANM consensus statements on PSMA PET/CT response assessment criteria (6), recommendations or guidelines on which segmentation algorithm or PSMA PET/CT–derived parameter(s) should be used are lacking. Various thresholding techniques, such as using fixed thresholds (mostly, SUV_max of greater than 3) or relative thresholds (e.g., 40%–45% of the SUV_max of the selected lesion), also exist for PET image segmentation (16–18,23). Here, we applied a combined fixed SUV_max of greater than 3 and a lesion volume threshold of greater than 0.5 mL to select and delineate PSMA-positive lesions. Although potential misinterpretation of background foci as small lesions was avoided in this way, this approach underestimated the number of liver metastases in 2 of 16 patients (12.5%) because of the difficulties in delineating lesions from the intense normal liver background activity. Combining liver-based and relative thresholds to limit image sampling errors and compensate for the spillover effect might also overcome the liver background–lesion discrimination issue (20,21). Moreover, as low-dose CT may underestimate small visceral lesions that can also be PSMA negative (24), PSMA imaging should be combined with thin-slice contrast-enhanced CT to optimize tumor burden enumeration and monitoring.

In contrast to tracer intensity of uptake, volumetric parameters were the most adequate for assessing treatment response using EAU/EANM PSMA PET/CT criteria in our dataset and the least influenced by the flare phenomenon (Supplemental Table 3). The underlying mechanism behind PSMA “flare” after androgen deprivation therapy is poorly understood. Similar to BS tumor flare definitions (25), the increase in SUV_max on ⁶⁸Ga-PSMA-11 PET/CT may also lead to a concomitant increase in PSMA-TV (and, thus, PSMA-TL) because of activity spillover or emergence of previously invisible or nonsignificant lesions at baseline and may result in the misinterpretation of PD, which is why the EAU/EANM did not recommend PSMA PET/CT imaging before 12 wk. The volumetric changes associated with a flare phenomenon may be significant but remain transitory; for example, for patient UPN17, the increase in SUV_max by 54% at W4 led to the appearance of 4 new lesions and an increase in PSMA-TL by 163%. By W12, the SUV_max had decreased by 70% (i.e., 16% lower than baseline), the previously observed new lesions disappeared completely, and PSMA-TL decreased by 49% from baseline (Fig. 1).

When comparing PSMA PET/CT at W4 and W12, we made 3 observations. First, an increase in SUV_max at W4 with a decrease in PSMA-TL, with or without new lesions, was confirmed at W12 to be linked to a flare phenomenon (e.g., for patients UPN2 and UPN14). Second, new lesions at W4 without an increase of greater than 30% in SUV_max, independently of PSMA-TL, were confirmed to be progressive at W12 (e.g., for patients UPN1 and UPN13). Third, when both SUV_max and PSMA-TL increased at W4, with or without new lesions, PD could not be distinguished from flare (e.g., for patients UPN12 and UPN17). Thus, defining PD on the basis of SUV_max alone does not seem to be feasible, and SUV_max should always be evaluated in combination with the other parameters to limit misinterpretation of flare as PD. Although at early time points SUV_max may provide a hint to a nuclear medicine specialist of the presence of a flare phenomenon, no flare was observed after W12, and SUV_max at W12 did not change the therapeutic response evaluation in our cohort.

Furthermore, the EAU/EANM recommendations on the use of uptake thresholds based on PERCIST were arbitrarily chosen, as these have been validated only for ¹⁸F-FDG PET. Even though tracer uptake in PSMA imaging does not reflect direct metabolic activity, modified PERCIST criteria were shown to perform better than morphologic criteria such as RECIST in metastatic PC—as molecular changes appeared earlier than morphologic ones (26). Although the aim of the present study was not to validate PERCIST criteria in PSMA imaging, we observed that caution should be taken when those criteria are used, especially for early imaging. Indeed, changes in tracer uptake are not synonymous with PD but rather seem to reflect biomolecular changes leading to modifications in PSMA expression, as indicated by the heterogeneous responses at the patient level, and further highlight the fact that additional data are needed to shed light on the mechanisms of PSMA expression and tracer uptake. Besides flare, the modulation of PSMA expression may also reflect intrinsic tumor tissue modifications conferring potential treatment resistance (10). In our data, the 5 of 16 patients (31%) with PD at W12 according to PSMA PET/CT had already met progression criteria at W4. Two of those patients (UPN12 and UPN13) had PD according to conventional imaging, and 1 patient (UPN16) had PSA progression.

With these EAU/EANM recommendations, patients with non-PD may be further subdivided into those with a stable response, those with a partial response, and those with a complete response, depending on the reductions in both SUV_max and PSMA-TL (Table 1). However, these criteria may need to be revised, as the extent of reduction in SUV_max and volumetric parameters rarely seemed comparable (Fig. 1). For example, at W12, 4 of 11 patients would be classified with a partial response (>30% reductions in both SUV_max and volumetric parameters) and 7 of 11 patients would be classified with a stable response even though 5 of the 7 achieved a significant (>30%) reduction in PSMA-TL. Data are also lacking on the thresholds that should be used, especially to define PD. For example, according to the current recommendations, PD may be defined by a 30% increase in tumor volume, but the recently proposed RECIP criteria have set a lower threshold of 20%; in addition, these parameters have been shown to carry prognostic value after ¹⁷⁷Lu-PSMA therapy (27). Moreover, in contrast to PERCIST, RECIP does not include tracer uptake modifications for evaluating response to ¹⁷⁷Lu-PSMA therapy. Nonetheless, this parameter could be of potential use for improving patient stratification before therapy initiation and was recently shown to predict a higher likelihood of a response to ¹⁷⁷Lu-PSMA therapy than to cabazitaxel (28).

The integration of minimally invasive molecular biomarkers, such as circulating tumor DNA, with novel imaging might facilitate discrimination between PD and flare and guide therapeutic intervention at early response assessment time points. As shown in a recent work, circulating tumor DNA does not seem to rise in patients with an increase in PSA or bone flare on conventional imaging (29). Additionally, the introduction of PSMA PET/CT in mCRPC might improve disease control rates by identifying oligoresistant or oligoprogressive lesions, which could be subjected to—for example—metastasis-directed therapy, while preserving the antitumoral effect of the systemic agent on the responsive lesions (12,13).

Overall, molecular imaging parameters have the potential to act as predictive biomarkers of response to treatment, but whether modifying a treatment plan according to them improves patient outcomes has yet to be determined in larger prospective trials. The main limitations of the present study were the small number of patients who were retrospectively included and the absence of validated criteria for the interpretation of PSMA PET/CT scans and the delineation method.