Diffuse infiltrating gliomas can have an astrocytic or oligodendroglial origin—World Health Organization (WHO) grades 2–4, depending on subtype—and have a high morbidity and mortality even with optimal treatment consisting of surgical resection and postoperative chemoradiotherapy (1). This is due to the (microscopic) infiltrative growth pattern of glioma and frequent treatment resistance, which in turn leads to frequently observed post-treatment tumor progression (TP), i.e., the renewed occurrence or progression of enhancing areas within the remaining tumor or surgical bed on follow-up conventional magnetic resonance imaging (MRI). However, treatment-related abnormalities (TRA), including pseudoprogression and radiation necrosis, have almost identical characteristics on conventional MRI (2), resulting in a diagnostic challenge (3, 4). Considering the fact that both entities require vastly different therapeutic approaches and are associated with significantly different outcomes, more sophisticated imaging techniques have been proposed to distinguish TP from TRA in recent years (5–7). These techniques include advanced MRI techniques (e.g., diffusion-weighted MRI, perfusion-weighted MRI, and MR spectroscopy) and positron emission tomography (PET) imaging with radiolabeled amino acids (AA-PET). Despite a histopathological evaluation of tissue obtained through a biopsy or resection being the gold-standard diagnostic approach, there is a medical need for highly accurate non-invasive techniques. Molecular imaging reveals the physiological properties of the tissue non-invasively and is therefore regarded as the most optimal non-invasive technique to differentiate TP from TRA (8). In general, it is thought that static PET imaging provides sufficient information when used in the postoperative setting in order to discern TP from TRA. However, static PET imaging does not capture all the biological properties of either TP or TRA lesions and dynamic imaging protocols have been developed, especially for O-(2-[¹⁸F]fluoroethyl)-L-tyrosine ([¹⁸F]FET)-PET imaging (9). The clinical value of such dynamic imaging protocols, in comparison with static imaging protocols, however, remains unclear. This study, therefore, investigated the diagnostic accuracy of static vs. dynamic AA-PET in distinguishing between TP and TRA in patients with post-treatment glioma.

In total, 33 patients [17 females, mean age of 55.7 ± 12.2 years (standard deviation)] were eligible for inclusion. Five patients were diagnosed with oligodendroglioma (WHO grade 2: 3; WHO grade 3: 2), seven with astrocytoma (WHO grade 2: 5; WHO grade 3: 2), and 21 with glioblastoma (WHO grade 4: 21) at baseline. Moreover, 13 PET-MRI examinations were carried out using [¹¹C]MET and 20 examinations were performed after the administration of [¹⁸F]FET. The mean administered doses of [¹¹C]MET and [¹⁸F]FET were 719 MBq (± 51 MBq) and 209 MBq (± 26 MBq), respectively. A more detailed overview of the included cohort according to tracer is provided in Table 1.

No significant differences in patient characteristics and PET parameters were found when comparing the two radiotracer groups (Table 1). This indicated that the static and dynamic variables derived from AA-PET were comparable for the neuro-oncological disorders covered in this study.

The TBR_mean and TBR_max values were significantly different between TRA and TP lesions when using [¹¹C]MET PET (p = 0.019 and p = 0.013, respectively). The analysis of the TTP in the dynamic [¹¹C]MET PET data revealed no significant differences between the TRA and TP groups (p = 0.007). Similarly, the analysis of the AUTAC derived from [¹¹C]MET PET data showed no significant differences between patients with TRA and those with TP (p = 0.121).

When comparing the TBR_mean values derived from the [¹⁸F]FET-PET data, there was no significant difference between the TRA and TP groups (p = 0.066). When comparing the TBR_max values derived from the [¹⁸F]FET-PET data between the TRA and TP groups, a significant difference was observed (p = 0.005). Furthermore, a significant difference was found in AUTAC values derived from the dynamic [¹⁸F]FET-PET data between the TP and TRA groups (p = 0.002). No significant differences were found between the TP and TRA groups when analyzing the TTP values derived from the dynamic [¹⁸F]FET-PET data (p = 0.271).

The analysis of the [¹¹C]MET PET data-derived parameters TBR_mean and TBR_max indicated in very-good-to-excellent diagnostic accuracy (AUROC = 0.93; 95%-CI: 0.73–1.00 and AUROC = 0.85; 95%-CI: 0.61–1.00, respectively) for the detection of TP. Moreover, the diagnostic accuracy of TTP was found to be very good (AUROC = 0.87; 95%-CI: 0.58–1.00), whereas the AUTAC demonstrated poor diagnostic accuracy (AUROC = 0.47; 95%-CI: 0.00–0.92). Based on the overlapping confidence intervals of the AUC values, none of the metrics was significantly better at distinguishing TP from TRA.

For TBR_mean, an optimal cut-off value of 1.83 provided a balanced sensitivity and specificity of 80% and 100%, respectively. For TBR_max, an optimal cut-off value of 2.3 provided a balanced sensitivity and specificity of 70% and 100%, respectively. Furthermore, for TTP, an optimal cut-off value of 141.5 provided a balanced sensitivity and specificity of 70% and 100%, respectively. Finally, for the AUTAC, an optimal cut-off value of 1,671.2 provided a balanced sensitivity and specificity of 60% and 67%, respectively.

The analysis of the [¹⁸F]FET PET data-derived parameters TBR_mean and TBR_max indicated a good-to-very-good diagnostic accuracy (AUROC = 0.73; 95%-CI: 0.39–0.98 and AUROC = 0.84; 95%-CI: 0.62–0.99, respectively) for the detection of TP. The diagnostic accuracy of TTP was found to be unsatisfactory (AUROC = 0.54; 95%-CI: 0.24–0.82), whereas the AUTAC demonstrated very good diagnostic accuracy (AUROC = 0.83; 95%-CI: 0.53–1.00). Based on the overlapping confidence intervals of the AUC values, none of the static or dynamic metrics was significantly better at distinguishing TP from TRA.

For TBR_mean, an optimal cut-off value of 2.36 provided a balanced sensitivity and specificity of 71% and 83%, respectively. For TBR_max, an optimal cut-off value of 3.3 provided a balanced sensitivity and specificity of 57% and 100%, respectively. Furthermore, for TTP, an optimal cut-off value of 127.5 provided a balanced sensitivity and specificity of 43% and 83%, respectively. Finally, for the AUTAC, an optimal cut-off value of 1,481.1 provided a balanced sensitivity and specificity of 93% and 83%, respectively.

Figure 3 shows the ROC curve for each of the static and dynamic PET parameters that were derived from either [¹⁸F]FET or [¹¹C]MET PET data for the differentiation between TP and TRA.

We found that dynamic AA-PET can provide some additional insights that can help distinguish TP from TRA in patients with post-treatment glioma. However, whether a 60-min dynamic AA-PET examination (especially when using [¹¹C]MET) is a necessity remains to be investigated in future studies. In general, a 20-minute static AA-PET examination seems to suffice to distinguish TP from TRA in clinical practice. This is contrary to the results reported by Kebir and colleagues, who found a mean time to peak of 25 min in TP vs. a mean time to peak of 40 min in TRA (P < 0.001) (13), as we found no differences in time to peak between patients with TP and TRA. Interestingly, our study shows a much earlier peak in both TP (2.15 min) and TRA (1.95 min) during dynamic AA-PET imaging with [¹⁸F]FET compared to Kebir et al. (13). This indicates that the vascular phase of the image acquisition plays an important role when using the state-of-the-art imaging equipment detailed in this study. The difference in results can be further explained by differences in the emission recording protocols. Kebir et al. reported the use of 16 time frames, in which the first 5 min are recorded in five time frames of 1 minute each, resulting in a coarser evaluation of early dynamics (13). In our study, however, greater emphasis was placed on the early dynamic phase. This emphasis is also recommended in the most recent international guidelines on this topic (9). This could help explain these findings, though more research is needed to investigate whether this hypothesis holds.

Furthermore, our study shows differences in the diagnostic capacity of TTP and AUTAC for [¹⁸F]FET and [¹¹C]MET PET imaging. Differences in the diagnostic capacity of dynamic PET imaging using [¹⁸F]FET and [¹¹C]MET are already well-known in the pre-treatment imaging setting. In this setting, only the use of [¹⁸F]FET has an established clinical value in dynamic PET imaging (5). Dynamic [¹⁸F]FET PET imaging has been reported to increase diagnostic accuracy in tumor grading, as WHO grade 2 gliomas typically show a steadily increasing time-activity curve, whereas gliomas graded as WHO grades 3 and 4 elicit an early activity peak around 10–20 min after injection, followed by a decrease of [¹⁸F]FET uptake (14, 15). Moreover, pre-treatment dynamic [¹⁸F]FET PET imaging has been proven to significantly increase the detection rate of more malignant foci (16, 17). However, the use of dynamic PET imaging using [¹¹C]MET in the pre-treatment setting did not result in improved diagnostic accuracy in tumor grading (18). To explain these contradictory findings, Moulin-Romsée et al. discussed the role of the system-L-transporter (LAT) proteins, which are overexpressed on the membrane of glioma cells. Although the mechanism of FET accumulation in glioma cells remains unclear, LAT subtypes 1, 2, and 3 are believed to play a role. Glioma cells overexpress LAT-1, which also allows the transport of smaller amino acids such as methionine, and is primarily driven by intracellular amino acid concentrations and composition (19). Since FET is not metabolized intracellularly (20, 21), the inward transport of FET is believed to exceed the outward transport via LAT-1. Methionine, in contrast, is incorporated into proteins in the glioma cell, which is hypothesized to halt the inward transport, explaining the pre-treatment imaging results (18). To some extent, these mechanisms of action may also help us understand our findings of differences in the diagnostic accuracy of [¹⁸F]FET and [¹¹C]MET PET imaging. Furthermore, it is known that differences in radiotracer distribution exist when comparing [¹⁸F]FET and [¹¹C]MET, including greater uptake of [¹¹C]MET in inflammatory lesions and different physiological uptake of [¹¹C]MET in normal brain tissue compared to [¹⁸F]FET (20, 22, 23). Another striking finding concerned the non-significant difference in the TBR_mean values derived from the [¹⁸F]FET PET data when comparing patients with TP and those with TRA (p = 0.066). This is not in line with the available evidence reported in the most recent guidelines (9) and must be considered a consequence of the relatively limited sample size in this specific sub-cohort.

This study had several limitations, including the retrospective, single-center design, which inherently introduces a risk of selection bias and limits its generalizability. Second, the relatively small sample size reduces statistical power, leading to wider confidence intervals and, possibly, instability of the subgroup analyses. Due to the previously mentioned differences in radiotracer distribution and differences in the subsequent metabolic processes between [¹⁸F]FET and [¹¹C]MET, no pooled analyses could be performed. Third, histopathology was not available in all patients and, although this is often the case in neuro-oncological research, this must be considered a limitation. Finally, dynamic PET protocols and reconstruction parameters may vary across centers, which could limit the generalizability of our findings. Despite these limitations, the results can provide clinically relevant guidance. The finding that static AA-PET parameters can be sufficient to differentiate TP from TRA implies that complex and time-consuming dynamic PET protocols may not be required in routine clinical practice. This can streamline imaging workflows, reduce patient burden, and facilitate timely therapeutic decision-making in neuro-oncology.

This study suggests that dynamic AA-PET parameters do not provide significant added value in differentiating TP from TRA in patients with post-treatment glioma. While static AA-PET parameters demonstrated good-to-excellent diagnostic accuracy and may suffice for clinical decision-making, the current analysis did not formally assess the incremental value of using dynamic metrics in addition to static measures. The small cohort size, retrospective design, heterogeneous reference standard, and potential segmentation/measurement variability may have limited the power of the study to detect any modest but clinically relevant improvements when using dynamic imaging. Therefore, while dynamic imaging demonstrated no additional diagnostic value, this does not rule out the possibility of such an effect in a larger or more optimized study.