Suspected recurrence of brain metastases after focused high dose radiotherapy: can [18F]FET- PET overcome diagnostic uncertainties?

Background After focused high dose radiotherapy of brain metastases, differentiation between tumor recurrence and radiation-induced lesions by conventional MRI is challenging. This study investigates the usefulness of dynamic O-(2-18F-Fluoroethyl)-L-Tyrosine positron emission tomography (18F-FET PET) in patients with MRI-based suspicion of tumor recurrence after focused high dose radiotherapy of brain metastases. Methods Twenty-two patients with 34 brain metastases (median age 61.9 years) were included. Due to follow-up scan evaluations after repeated treatment in a subset of patients, a total of 50 lesions with MRI-based suspicion of tumor recurrence after focused high dose radiotherapy could be evaluated. 18F-FET PET analysis included the assessment of maximum and mean tumor-to-background ratio (TBRmax and TBRmean) and analysis of time-activity-curves (TAC; increasing vs. decreasing) including minimal time-to-peak (TTPmin). PET parameters were correlated with histological findings and radiological-clinical follow-up evaluation. Results Tumor recurrence was found in 21/50 cases (15/21 verified by histology, 6/21 by radiological-clinical follow-up) and radiation-induced changes in 29/50 cases (5/29 verified by histology, 24/29 by radiological-clinical follow-up). Median clinical-radiological follow-up was 28.3 months (range 4.2–99.1 months). 18F-FET uptake was higher in tumor recurrence compared to radiation-induced changes (TBRmax 2.9 vs. 2.0, p < 0.001; TBRmean 2.2 vs. 1.7, p < 0.001). Receiver-operating-characteristic (ROC) curve analysis revealed optimal cut-off values of 2.15 for TBRmax and 1.95 for TBRmean (sensitivity 86 %, specificity 79 %). Increasing TACs and long TTPmin were associated with radiation-induced changes, decreasing TACs with tumor recurrence (p = 0.01). By combination of TBR and TACs, sensitivity and specificity could be increased to 93 and 84 %. Conclusions In patients with MRI-suspected tumor recurrence after focused high dose radiotherapy, 18F-FET PET has a high sensitivity and specificity for the differentiation of vital tumor tissue and radiation-induced lesions.


Background
Focused high dose radiotherapy, such as stereotactic radiosurgery (SRS), is increasingly used for small sized brain metastases especially in the "oligometastases situation" [1,2]. Low-activity iodine-125 brachytherapy (SBT) is another ablative strategy which has been shown to be effective especially in recurrent brain metastases after previously performed radiotherapy [3].
The treatment rationale of both SRS and SBT is to obtain tumor control through early cytocidal effects as well as late vascular changes [4]. However, lesions that undergo such treatment constitute a diagnostic challenge: in conventional follow-up magnetic resonance imaging (MRI) a new ring-shaped contrast enhancement can arise at the site of the highest delivered dose as indication of blood-brain barrier (BBB) disruption. These lesions will either expand or resolve over time [5][6][7]. Therefore, conventional MRI cannot adequately distinguish between tumor recurrence and treatment response [8]. Recent data suggest that molecular imaging techniques might help to overcome such limitations in intracranial metastases which had previously undergone focused high dose radiotherapy. Whereas positron emission tomography (PET) with the widely used 18 F-2deoxy-2-fluoro-D-glucose ( 18 F-FDG) has low diagnostic accuracy after SRS, the use of radiolabelled amino acids or amino acid analogues such as L-methyl-11 C-methionine ( 11 C-MET) and O-(2-18 F-Fluoroethyl)-L-Tyrosine ( 18 F-FET) reaches sensitivity and specificity values in the range of 78 and 100 % rendering especially 18 F-FET a highly reliable tracer in glioma imaging [9][10][11][12][13][14].
Still, there is only limited data on the diagnostic impact of 18 F-FET PET in intracranial metastases, especially in case of previous exclusive focused high dose radiotherapy [12,15]. For further clarification, we objected to evaluate if 18 F-FET PET can differentiate adequately between vital tumor recurrence and radiationinduced lesions.

Materials and methods
Patients Twenty-two patients (median age 61.9 years; 11female/ 11 male) with a suspicion of a recurrence of their brain metastases after previously performed focused high dose radiotherapy (i.e. SRS and SBT) were included. To rule out potentially confounding effects of WBRT in addition to focused high dose radiotherapy on 18 F-FET uptake, we excluded all patients with a history of whole brain radiotherapy. Thirteen patients suffered from of a single tumor and 9 patients from multiple metastases (3 lesions in 3 patients, 2 lesions in 6 patients), which resulted in 34 evaluable lesions.
Static and dynamic 18 F-FET PET was performed in all these patients at the time of the suspected tumor recurrence. The results of 18 F-FET PET investigation and their classification as either concordant or discordant to the actual tumor status referred either to the corresponding tissue diagnosis obtained from PET-guided biopsy procedures at the time of suspected recurrence and/or to the results of clinical/radiological follow up evaluation. In patients with deteriorating neurological status and/or further lesion growth with steroid resistant edema after high dose radiotherapy histological evaluation was aimed whenever possible. Lesions exhibiting unchanged or regressive MRI findings and a stable/improved clinical performance over a time period of ≥ 6 months after the date of the suspected tumor recurrence were classified as stable disease. When there were new suspicious imaging changes of a lesion and/or clinical deterioration after a ≥ 6 months time interval of stable/regressive disease after the initial treatment, this suspicion was handled as new case with subsequent new 18 F-FET PET investigation and new follow-up. Each individual metastasis per patient was evaluated. Criteria for recurrence were based on the Macdonald criteria [16]: i) the appearance of a new contrast enhancing area after previously observed complete response or ii) an increase of the enhancing area (>25 %) after previously observed partial response/stable disease. Accordingly, an overall number of 16 metastases (8 patients) were repeatedly analysed at different stages of the brain disease. Seven out of these 16 tumors underwent re-treatment with focused high dose radiotherapy (time interval between first and second treatment: 25 months). In the remaining nine lesions, 18 F-FET PET re-evaluation was initiated due to an increase in contrast enhancement and/or steroid-resistant symptomatic edema after a median latency period of 8.6 months. Of these 9 lesions, 6 were found to be tumor recurrences and 3 were found to be radiation induced lesions (median follow up 23.1 months; range 13.6-33.9 months). Thus, a total number of 50 brain lesions was analysed.
Clinical parameters were determined using patients' electronic medical records and paper charts. Patients were evaluated and consented by both an experienced neurooncologist (FWK) and nuclear medicine physician (NLA). The study was approved by the institutional review board. All patients had given written informed consent.

Radiosurgery, iodine-125 stereotactic brachytherapy and stereotactic biopsy
Three-dimensional planning for SRS and SBT was routinely based on computerized tomography (CT) and MRI (slice thickness <3 mm). The treatment protocol for SRS was used as described in Schüttrumpf et al. [2]. In brief, the clinical target volume was set equivalent to the gross tumor volume. Expansion of the gross tumor volume with 1-2 mm resulted in the planning target volume. Radiation dose ranged from 18 to 24 Gy. For SBT, the definition of the treatment volume and the treatment planning was done as described in Schwartz et al. using the BrainLab AG Target software (version 1.19) [17]. The prescribed reference dose calculated to the outer tumor rim was generally 54.0 Gy. The dose rate was < 15.0 cGy/h. Prior to SBT, the diagnosis of a metastasis was confirmed with stereotactic biopsy as previously described [18]. Whenever the attending neuropathologist could make the diagnosis of a vital metastasis using smear specimen intraoperatively, SBT was carried out immediately. In case of intraoperative uncertainty regarding the correct diagnosis, SBT was withheld and the results of the paraffin embedded analysis was awaited (all sections were stained with hematoxylin and eosin). Another CT scan was done at the first postoperative day to demonstrate the correct position of the implanted seed(s). 18

F-FET PET image acquisition and evaluation
Dynamic 18 F-FET PET image acquisition was performed over 40 min after injection of 180 MBq 18 F-FET as described previously at a Siemens ECAT EXACT HR+ PET scanner [6,19]. Image data were transferred to a HERMES workstation. 18 F-FET PET analysis was performed for each lesion and included the assessment of the maximum standardized uptake value (SUV) as tumor-to-background ratio (TBR max ), the mean tumorto-background ratio (TBR mean ) and dynamic analysis of tumoral 18 F-FET uptake over time (increasing vs. decreasing time-activity-curves (TACs)). For the determination of background activity, a region-of-interest was drawn in six subsequent slices in the contralateral hemisphere including grey and white matter, added to a volume-of-interest, and the mean SUV was set as background acitivity. The TBR max and TBR mean were calculated by using the maximum and mean SUV within a semiautomatically drawn, threshold-based volume-ofinterest (1.8 x background activity). For the dynamic analysis, a 90 % isocontour threshold region-of-interest was drawn in the 10-30 min summation images on each slice of suspicious uptake, and the respective TACs for the whole dataset were extracted. TACs within a 40 min time interval after tracer injection were defined as follows: (i) lesions with homogeneously increasing TACs with SUV constantly ascending or reaching a peak followed by a plateau in the subsequent frames was classified as having increasing TACs and (ii) lesions with homogeneously or heterogeneously decreasing TACs with SUV showing an early peak within the first 20 min time interval followed by a constant descent thereafter were classified as having decreasing TACs. Early fluctuations in the TACs within the first short time frames (7 × 10 s followed by 3 × 30 s) representing noise were excluded from kinetic analyses. For each slice within the tumor, the frame with the peak uptake was identified. The starting time of the frame plus half the frame duration, corresponding to the respective peak value, was set as time to peak (TTP). The shortest TTP present in at least 2 adjacent slices was defined as minimal TTP (TTP min ) as described in Jansen et al. [20].
PET findings were evaluated by consensus reading of two experienced nuclear medicine physicians (NLA and MU) who were blinded for clinical and histological data as previously described [21].

MR imaging
Clinical evaluations were combined with MRI at regular 3-month intervals. MRI protocols consisted of T1weighted ± gadolinium contrast medium, T2-weighted and fluid attenuated inversion recovery (FLAIR) sequences. Slice thickness was 1.0 mm for all MRI sequences and all sequences were reconstructed in axial, coronal and sagittal planes.

Statistical analysis
SPSS for Windows (SPSS, Version 21.0, Chicago, IL) was used for statistical calculations. The evaluated PET parameters (TBR max and TBR mean and increasing vs. decreasing time activity curves) were correlated with histopathological results when available and with the clinical follow-up. The comparison between tumor recurrence and radiation-induced lesion was performed using the Mann-Whitney-U test for continuously scaled variables and using the χ 2 test for categorical variables. Continuous parameters were reported as mean ± standard deviation and range. A two-tailed p-value < 0.05 was considered significant. Receiver operating characteristic (ROC) analyses were performed in order to identify the optimal TBR max cut-off value for the discrimination between tumor recurrence and radiation-induced lesion by choosing the cut-off leading to the highest product of sensitivity × specificity. Length of local PFS was calculated from the date of radiosurgery and analysed with the Kaplan-Meier method.

Results
Patient characteristics and 18 F-FET PET results are summarized in Table 1. Tumor recurrence was found in 21/50 cases (15/21 verified by histology, 6/21 by radiological-clinical follow-up) and radiation-induced lesions in 29/50 cases (5/29 verified by histology, 24/29 by radiological-clinical follow up data are included in Table 1). Median clinical and radiological follow up was 28.3 months (range 4.2-99.1 months; individual follow up data are indicated in Table 1), one patient was lost to follow-up. Median time from focused high dose radiotherapy to 18 F-FET PET was 13.7 months (range 1.9-   According to this cut-off value, 18 F-FET PET was false negative in 6 % of cases and false positive in 12 % of cases. In the false negative cases (n = 3), a median TBR max of 1.9 (range 1.5-2.1) and a TBR mean of 1.7 (range, 1.4-1.9) were found, median time from radiotherapy to 18 F-FET PET was 4.1 months (range, 2.4-5.3 months). In the false positive cases (n = 6), a median TBR max of 3.1 (range 2.6-3.5) and a TBR mean of 2.2 (range 2.0-2.4) were seen (median time from radiotherapy to 18 F-FET PET 10.4 months (range 2.5-23.5 months)).

Evaluation of time activity curve patterns and TTP min
Analysis of TAC patterns was performed in lesions with increased 18 F-FET uptake (TBR max ≥ 1.6) and was available in 42/50 cases. Increasing TACs were seen in 18 cases which were mainly associated with radiation-induced lesions (14/18 cases, 78 %), while decreasing TACs were found in 24 cases, of which 16 (67 %) were tumor recurrences (p = 0.01; two illustrative cases are given in Fig. 1 and 2). TTP min was significantly shorter in tumor recurrences (median 17.5 min) than in radiation-induced lesions (median 35 min; p = 0.007). Although most of the cases with short TTP min ≤ 12.5 min had a tumor recurrence (9/12 cases, 75 %) and most of cases with late TTP min of 35 min were radiation-induced (12/15 cases, 80 %), no reliable cut-off TTP min could be defined, as the rate of tumor recurrence and radiationinduced lesion in lesions with an intermediate TTP min of 17.5-25 min was 53.3 % vs. 46.7 %. Sensitivity for the detection of tumor recurrence by the mere qualitative classification using decreasing TACs as indicator for tumor recurrence was 80 %, specificity was 63 % (accuracy 71 %, AUC 0.72, CI 0.55-0.88, p = 0.02). By adding dynamic to static PET information, the sensitivity and specificity for the detection of tumor recurrence could be increased: for lesions with TBR max > 2.15 / TBR mean > 1.95 in combination with decreasing TACs a sensitivity of 93 % and specificity of 83 % were obtained (accuracy 87 %, AUC 0.79, CI 0.66-0.92, p = 0.001).
Static and dynamic PET findings are summarized in Tables 1 and 2. Nearly identical results could be achieved for the subgroup of the 34 lesions undergoing a single 18 F-FET PET evaluation (data not shown).

Discussion
Differentiation between radiation-induced lesions and tumor recurrence after focused high dose radiotherapy of brain metastases is challenging [8,15]. In the current series, nearly 60 % of the metastases exhibiting clinical and/or radiological signs of tumor progression turned out to be radiation-induced lesions. In this context, the place of metabolic imaging and suitable radiotracers still needs to be defined. 11 C-MET, 18 F-DOPA and 18 F-FET PET seem to be promising candidates for clinical routine [10,11,22,23]. The use of 11 C-MET however, is limited by logistic disadvantages (half life of 20 min restricting the use to sites with an on-site cyclotron [11]. Regarding Our results are in line with this study. This is especially true for the congruent TBR mean cut-off value of >1.95, rendering this parameter clinically valuable due to its applicability in different sites. Interestingly, the TBR mean threshold was identical despite differences in pretreatment between the two studies (exclusive focused high dose radiotherapy in our study versus mixed pretreatment with whole brain radiotherapy in the study by Galldiks). This might indicate that focused high dose radiotherapy itself determines the threshold values in PET imaging and that additional pretreatment with whole brain radiotherapy might not have a profound impact on TBR values.
Of note, compared to the above-mentioned study by Galldiks and colleagues, a lower cut-off value for TBR max was found in our series (2.15 vs. 2.55). This might be explained by differing image reconstruction protocols, i.e. filtered-back projection in our study versus iterative reconstruction in the above-mentioned study. Lately, iterative reconstruction has been reported to be associated with higher TBR max values in gliomas compared to filtered-back projection [24]. Therefore, the reported TBR max cutoff values can possibly only be used with the same image reconstruction parameters, while the TBR mean cut-off value seems to be more stable.
Our findings uncover diagnostic pitfalls after focused high dose radiotherapy of brain metastases. Our false positive cases (6/50) had markedly high TBR values (median TBR max 3.1, range 2.6-3.5; median TBR mean 2.2, range 2.0-2.4). Pretreatment was SRS in 4 cases and SBT in 2 cases and latency between 18 F-FET PET and treatment ranged from 2.5-23.5 months (median 10.4 months). With regard to these parameters, the false positive cases did not differ from the correctly diagnosed tumor recurrences so that no explanation can be found for the apparently unspecific high 18 F-FET uptake in these radiation-induced lesions. In our false negatives cases (3/50), the lesion size was remarkably small (median MR volume 0.6 ml, range 0.2-1.1 ml). In these cases, both the partial volume effect and the resolution have possibly contributed to the low TBR max and TBR mean values which were below the cutoff and therefore did not identify these small lesions as tumor tissue. In summary, these false positive/false negative result might anticipate a "lesional instability" leading to following clinical implications for the management of patients with brain metastases after focused high dose radiotherapy: i) suspicious lesions with high TBR values exceeding the cutoff should be histologically evaluated to verify tumor recurrence and to avoid overtreatment; ii) suspicious lesions with low TBR values below the cutoff do not need to be evaluated histologically and can be observed, however, iii) when being of small volume, should undergo closer follow-up as partial volume effects might hamper the PET evaluation. Furthermore, our observed rate of radiation-induced lesions as function of time implies that the risk for a radionecrosis is highest within the first two years after high-dose radiotherapy. Therefore, the time interval between radiotherapy and suspicion of tumor relapse should be taken into account in the further treatment rationale especially within the first two years. Eventhough TTP min was significantly shorter in tumor recurrences as compared to radiation-induced lesions, no reliable cutoff could be established for this parameter currently diminishing its prognostic impact. More data are necessary to determine its place within the diagnostic platform of brain metastases.
It should be stressed that not all patients had undergone stereotactic biopsy for histopathological verification (this was considered unethical in patients with suspected stable radiation-induced lesions). Systematic evaluation of 18 F-FET PET uptake patterns in untreated and already irradiated tumors under consideration of the primary tumor will certainly help to overcome data heterogeneity and might be realized in the framework of future multicentre trials. The still increasing number of cancer patients will act as a pressure into this direction.

18
F-FET PET appears to be an attractive tool in the differentiation of tumor recurrences and radiation-induced lesions and its place is clearly beyond structural imaging. The combination of static (TBRs) and dynamic (TACs) parameters shows high sensitivity, specificity and accuracy values. Still, 18 F-FET PET should be considered an addition rather than a replacement of stereotactic biopsy and close follow-up. This is especially true for selected ambiguous cases to avoid overtreatment or undertreatment.