Added value of [18F]FDOPA PET to the management of high-grade glioma patients after their initial treatment: a prospective multicentre study

Diagnostic value of 3,4-dihydroxy-6-[18F]fluoro-L-phenylalanine ([18F]FDOPA) PET in patients with suspected recurrent gliomas is recognised. We conducted a multicentre prospective study to assess its added value in the practical management of patients suspected of recurrence of high grade gliomas (HGG). Patients with a proven HGG (WHO grade III and IV) were referred to the multidisciplinary neuro-oncology board (MNOB) during their follow-up after initial standard of care treatment and when MRI findings were not fully conclusive. Each case was discussed in 2 steps. For step 1, a diagnosis and a management proposal were made only based on the clinical and the MRI data. For step 2, the same process was repeated taking the [18F]FDOPA PET results into consideration. A level of confidence for the decisions was assigned to each step. Changes in diagnosis and management induced by [18F]FDOPA PET information were measured. When unchanged, the difference in the confidence of the decisions were assessed. The diagnostic performances of each step were measured. 107 patients underwent a total of 138 MNOB assessments. The proposed diagnosis changed between step 1 and step 2 in 37 cases (26.8%) and the proposed management changed in 31 cases (22.5%). When the management did not change, the confidence in the MNOB final decision was increased in 87 cases (81.3%). Step 1 had a sensitivity, specificity and accuracy of 83%, 58% and 66% and step 2, 86%, 64% and 71% respectively. [18F]FDOPA PET adds significant information for the follow-up of HGG patients in clinical practice. When MRI findings are not straightforward, it can change the management for more than 20% of the patients and increases the confidence level of the multidisciplinary board decisions.


Introduction
High grade gliomas are the most frequent malignant brain tumours in adults [1]. There is a consensus on the initial treatment based on surgery, radiotherapy and chemotherapy. Unfortunately, the responses to this standard-of-care are not durable and additional treatments are necessary and can involve re-irradiation, alternative chemotherapy, clinical trials for innovative drugs and/or repeat resection. Standardof-care in this setting of recurrence is far less well defined [2] and the prognosis remains poor with rare long term survivals [3].
During follow up of initially treated patients and when recurrence is suspected, individual decisions about continuation, modification or discontinuation of treatment are usually based on magnetic resonance imaging (MRI). However, the initial therapeutic strategies, besides their effectiveness on the tumours, also induce effects on the surrounding tissues. These post-treatment changes are responsible for complex MRI modifications, in particular on contrast enhancement 1 3 which was the core of the initial structural Macdonald criteria [4]. Therefore, new criteria have been proposed to cope with these phenomena [5]. Yet, it is clear that structural imaging alone is limited in that setting and that advanced multimodal imaging is important to improve the non-invasive characterisation of post-therapeutic changes in brain tumours [6,7]. Advanced MRI techniques including perfusion weighted sequences, diffusion weighted imaging and spectroscopy are very helpful but still suffer from a lack of standardization [8][9][10]. MRI remains the corner stone of high-grade glioma patients' follow-up despite these difficulties, and additional work is needed to improve response assessment [11].
Nuclear medicine molecular imaging using positron emission tomography (PET) and various radiopharmaceuticals is able to provide additional information such as local glucose consumption or amino acid uptake. Evidence based recommendations for the use of PET imaging in the management of glioma patients have been proposed by Albert et al. [12]. In the setting of post-treatment evaluation, [ 18 F] fluoro-deoxy-glucose has been found to be of moderate additional value to MRI for differentiating progression or recurrence versus treatment induced changes. Conversely, amino acid PET has demonstrated a high sensitivity and specificity for the differentiation between progression and treatment-related changes. This is true for the 3 most widely used labelled amino acid and analogues: These are particularly attractive for imaging brain tumours because of their high accumulation in tumour tissues and their low uptake in normal brain. The increase of amino acids in brain tumours is related to the over-expression of the amino-acid transporters (LAT) in tumour cells and their vasculature [13,14]. Since LAT are normally expressed at the blood brain barrier, these compounds do not need blood brain barrier breakdown like contrast media to be untaken by brain lesions. The use of [ 11  F]FDOPA is the physiological striatal uptake for the latter. This normal striatal uptake of [ 18 F]FDOPA represents a significant advantage since it gives an internal normal reference. Its drawback is the potential difficulty to separate this activity from tumour uptake when the tumour is adjacent to striatum.
[ 18 F]FDOPA PET has demonstrated a higher accuracy than MRI in the differentiation of glioma recurrence from treatment-induced changes [12]. However, in practice, MRI is systematically performed first and [ 18 F]FDOPA PET is secondarily used in association only in some cases. Its added value ultimately lies in its capability of changing treatment decisions [15]. Therefore, prospective studies assessing its impact on patients' management in those conditions are necessary.
Two monocentric studies have addressed this issue. Walter et al. [16] first studied 58 patients with high and low grade brain tumours in different settings and found that [ 18 F] FDOPA PET changed the intended management in 41% of cases. However, the study did not include MRI in the decision process. More recently Humbert et al. [17] studied the impact of the [ 18 F]FDOPA PET results along with MRI, on a multidisciplinary brain tumour board decisions. They studied 56 glioblastoma patients and 41 with brain metastases in the setting of residual disease and recurrence. Treatment plans were changed in 33.3% of the glioblastoma patients and in 17% of the metastatic ones. However, multicentre studies were lacking so far [12].
We conducted a multicentre prospective study to assess the added value of [ 18 F]FDOPA PET to conventional clinical and MRI based decision making in the specific setting of post-treatment evaluation of patients with high grade gliomas in daily practice, namely in the differential diagnosis between recurrence/progression and treatment-induced changes.
The objective was to measure the changes in diagnosis and patient's management proposed by the Multidisciplinary Neuro-Oncology Board (MNOB) when taking into account the [ 18 F]FDOPA PET information.

Patient selection criteria
This was a multicentre, open, uncontrolled and non-randomized study designed to evaluate the added value au [ 18 F] FDOPA to the conventional management of high grade glioma patients.
Inclusion criteria were (i) histopathologically proven high grade gliomas (WHO grade III and IV); (ii) age over 18 years; (iii) signed informed consent; (iv) patients referred to the MNOB of each institution in the setting of decision making in the follow-up after initial standard of care treatment (surgery and radio-chemotherapy) and (v) patients for whom MRI findings were not straightforward.

3
MRI findings were considered not straightforward when the RANO criteria [5] were not leading to a sufficient confidence on the MNOB diagnosis.

Study design (Fig. 1)
Each patient's case was presented and discussed during the weekly meetings of the MNOB of each institution. The MNOB included neuro-oncologists, neurosurgeons, radiologists, pathologists and radiotherapists. The presence of a nuclear physicians was not mandatory.
The review process systematically included 2 steps. During step 1, the MNOB discussion was only based on the clinical data and the MRI findings (including all available sequences) without knowledge of the [ 18 F]FDOPA PET results. This discussion led to a first diagnostic and a first management proposal. The possible diagnostic proposals were: complete response (CR), partial response (PR), stable disease (SD) or progressive disease (PD). The possible management proposals were: continuation of the same treatment modality or change of treatment modality. Changes in treatment modalities included, surgery, re-irradiation, different chemotherapy or end of curative treatments (supportive care). The confidence level of this first proposal was assessed using a 3 level confidence score.
During step 2, the same case was discussed again by the MNOB, this time including the [ 18 F]FDOPA PET images. With this additional information, a second set of diagnostic and management proposals was made. The same proposals as in step 1 were possible and a new confidence level was attributed.
The step 1 and step 2 proposals were compared and when unchanged, the levels of confidence were taken into account.
Patients had a systematic clinical follow-up at 3 months after the MNOB decision and exited the study after 12 months of total follow-up. Clinical data and, if surgery occurred, pathological data were collected.
Patients could undergo several MNOB reviews during the follow-up period.
For the evaluation of the diagnostic performances, results were compared to pathology when available or to the radiological and clinical follow-up. The radiological and clinical data used as diagnostic reference were those collected during the systematic 3 months visit after the MNOB decision.
For analysis by the MNOB, PET images were co-registered to post-contrast T1-weighted MR images obtained within 28 days from the PET study. Other MRI sequences were used for the first diagnosis and management procedure including T2-weighted FLAIR. Diffusion and perfusion imaging as well as proton-spectroscopy were not performed systematically but used when available.
PET images were displayed using a colour scale for which the maximum was set on the striatal region. The criteria for [ 18 F]FDOPA PET analysis were only visual according to   [19,20]: 0, no detectable lesion uptake; 1, detectable lesion uptake but less than striatum uptake; 2, lesion uptake equivalent to striatum uptake and 3, lesion uptake greater than striatum uptake. [ 18 F]FDOPA PET was considered positive, thus in favour of recurrence or progression, if the visual score was equal or greater than 2. This visual reading was done collectively during the MNOB meeting in each centre.
A post-hoc semi-quantitative analysis was performed on the centralized data (135 [ 18 F]FDOPA PET studies). Tumour to striatum ratios (TSR) and tumour to brain ratios (TBR) were measured using 3D 15 cc VOIs. The SUV peak values were used for the striatum and for the tumour. The SUV mean value measured within a mirrored VOI avoiding the striatum was used for the brain region.

Statistical analysis
The required number of patient cases to be evaluated was calculated prior to the start of the study to be 110 (considering 10% of patients could be wrongly included, lost to follow-up or withdrawn from the study). Data entry and management were performed on the capture system (Ennov Clinical). Categorical data are shown as counts and percentages and continuous variables as means with standard deviations. Comparisons were made using McNemar test for categorical paired data. Statistical analyses were performed with SAS software version 9.4 (SAS Institute Inc) and R 3.6.0 (R Foundation) on Windows® and the caret and DTComPair packages. Results were considered statistically significant at a p value of less than 0.05 (two-sided).

Patient population
One hundred and ten patients were included. Three of them did not meet the inclusion criteria (Fig. 2) leading to a studied population of 107 patients (62 males and 45 females) of median age 56 years (range 19-83). Patients' characteristics are given in Table 1. Ninety-one patients (85%) had a glioblastoma and 16 a WHO grade III glioma (15%). IDH1 status was mutated in 13 patients (12.1%), wild-type in 74 patients (69.2%) and NOS in 20 patients (18.7%).
In the majority of cases, patients were studied after having undergone surgery and received radiotherapy with concurrent adjuvant temozolomide (Stupp) (89 patients; 83.2%).

Impact of [ 18 F]FDOPA PET on Multidisciplinary Neuro-Oncology Board decisions
The 107 patients underwent a total of 138 [ 18 F]FDOPA PET studies leading to 138 MNOB assessments (Fig. 2 Table 2). The proposed management between step 1 and step 2 (without and with [ 18 F]FDOPA PET data) was unchanged in 107 cases (77.5%) and changed in 31 cases (22.5%). [ 18 F] FDOPA PET data led the MNOB to change its management proposal in favour of introducing a new chemotherapy in 14 cases, continuing the same treatment in 10 cases, performing surgery in 3 cases, radiotherapy in 2 cases and to stop all curative treatment in 2 cases (Table 3).

Diagnostic accuracy of each step
Considering the results of surgery in seven cases and the 3 months radiological and clinical follow-up for the rest as gold standard, step one of the process, only including clinical and MR data, diagnosed PD (progression/recurrence) with a sensitivity of 83%

Semiquantitative analysis
The AUROCs were 0.64 and 0.65 for the TSR and TBR respectively. With a best threshold of 1.03 for the TSR and of 1.94 for the TBR; their sensitivity, specificity and accuracy were 90%, 58%, 70% and 92%, 57%, 70% respectively.

Discussion
To our knowledge, this is the first prospective multicentre study of the impact of [ 18 F]FDOPA PET on the clinical management of high-grade gliomas in the setting of suspicion of recurrence after initial treatment. In the framework of effectiveness of diagnostic imaging [21], it corresponds to a level 3 type study of "diagnostic efficacy thinking" assessing the post-test changes induced by [ 18 F] FDOPA PET imaging. It shows that [ 18 F]FDOPA PET changed the MRI based intended management in 31 cases (22.5%) and increased the confidence in the initial decision when unchanged in 87 cases (81.3%).
Walter et al. [16] first studied 58 patients with high and low grade brain tumours in different settings and found that [ 18 F]FDOPA PET changed the intended management in 41% and the management changes were actually implemented in 31% of cases. Unlike in our study, PET data were not  However, the majority of the patients were studied in the context of suspected recurrence and MRI changes; the reported impact on the intended management in this subgroup was also 41%. This study included 38% of non WHO III or IV grades. The specific impact in low versus highgrade brain tumour management or in initial versus followup setting is unknown. This higher impact of [ 18 F]FDOPA PET is probably mostly due to the study design based on a 3 questionnaires survey sent to the referring physician in Walter's study whereas in our case these changes were decided on multidisciplinary bases in a directly operational setting. Furthermore, the final diagnostic classification in CR, PR, SD or PD in our case rather than a binary assessment might have contributed to reducing the number of decision changes. Walter [17], addressed a similar issue using [ 18 F] FDOPA PET. The design of the study was similar to ours, however this was a prospective monocentric study which included brain metastases as well as high grade gliomas during the follow up after initial treatment. Furthermore, within the high-grade glioma patients, only 12 were studied for the diagnosis of tumour recurrence. In that subgroup, the management was changed in one third of the cases. Our study found a smaller percentage of 22.5%, which is likely to be more realistic since obtained on 138 studied cases studied in 5 different institutions.
Other studies addressed a similar issue using other radiopharmaceuticals. Hillner et al. [22] investigated a large series of 367 patients with primary brain tumours from the National Oncological PET registry who underwent [ 18 F] FDG brain PET. They analysed pre and post PET forms Fig. 3 Examples. Top row: a 63 yo female patient with a glioblastoma treated by surgery and radio-chemotherapy 2 years prior to the study. The proposed diagnosis and management by the MNOB with MRI alone was: complete response and proposition of follow-up (medium confidence score). [ 18 F]FDOPA shows a significant uptake in the tumour site (score 3; higher than striatum). The proposed diagnosis and management by the MNOB including [ 18 F]FDOPA PET data was changed to progressive disease and proposition of surgery or radiosurgery (high confidence score). Bottom row: a 63 yo female patient with a non-resectable glioblastoma treated by radio-chemotherapy ending 2 month before the study. The proposed diagnosis and management by the MNOB with MRI alone was: stable disease and proposition of chemotherapy (medium confidence score). [ 18 F] FDOPA tumour site uptake was low (score 1; lower than striatum). The proposed diagnosis and management by the MNOB including [. 18 F]FDOPA PET data was changed to partial response and proposition of follow-up (high confidence score) filled by the referring physician and found 38.2% of intents of management changes in light of the [ 18 F]FDG PET findings. However, the paper could not document if the planned management changes were actually completed.
Yamane et al. [23] used [ 11 C]methionine PET to separate recurrence from radiation necrosis in brain tumours and its clinical impact based on retrospective questionnaires to the referring physicians. Twenty PET studies were performed for initial diagnosis of brain tumours and 69 for differentiating tumour recurrence from radiation necrosis. In this last subgroup which included various grades of primary brain tumours and metastases, intended management was confirmed in 42 cases and a significant management change due to PET results was found in 18 cases (42.9%).
Overall, our study performed prospectively in 5 different centres, shows a slightly inferior rate of changes of patients' management induced by [ 18 F]FDOPA PET than reported in the literature. However, this is more likely to be generalizable. By design, [ 18 F]FDOPA PET was performed only when MRI findings were not leading to a high confidence level in the decision to be made by the MNOB. This option was taken in order to be as close as possible to a realistic setting. Therefore, straightforward patients were not included and the study considered only equivocal cases. This can contribute to this discrepancy.
Furthermore, our study also shows that, when the decision itself was not modified, [ 18 F]FDOPA PET results increased the confidence of the decision in more than 80% of the cases.
In the literature, the performances of [ 18 F]FDOPA PET are generally reported as superior to MRI for the diagnosis of HGG recurrence [12,24,25]. This is also the case in our study. However, it should be pointed out that we measured the performances of step 2 process. Therefore [ 18 F]FDOPA PET performances were not assessed independently from those of MRI since both information were taken into consideration at this step. This corresponds to routine practice in which 18F-FDOPA is usually not interpreted without the MRI findings. However, we observe a sensitivity and a specificity of 84% and 57% which are rather in the lower range of the generally reported values for [ 18 F]FDOPA PET alone in that indication [19,[24][25][26] probably reflecting the multicentre design of our study as well as the relatively more complex cases included.
Decisions were taken based on a visual consensual analysis during the local MNOBs. The scores given by the boards were compared retrospectively to a blind central review by 3 nuclear physicians. There was a good concordance for separating abnormal from normal uptake (score equal of superior to 2) (kappa = 0.77). This shows that this visual score can be used within a multidisciplinary setting.
Semiquantification performed centrally using TSR and TBR did not provide an accuracy superior to visual reading (70% versus 71% respectively) as previously shown in another setting [19]. Dynamic analysis has not yet demonstrated a significant benefit in that setting [28,29].
An important limitation of the study concerns the diagnostic performances that were evaluated on surgical confirmation in only 7 cases and on the 3 month follow-up findings for the rest. However, this reflects the daily clinical practice in which few patients benefit from a repeat resection in case of recurrence [27]. Patients were included when MRI was not straightforward. Despite the application of the RANO criteria, it is likely to have introduced some variability of patients' inclusions among the centres. Lastly, the study evaluated changes in the management induced by the [ 18 F]FDOPA PET without assessing the impact on patient survival. A prospective study comparing the prognosis and the quality of life of patients who benefited from [ 18 F]FDOPA PET to patients treated without this additional information is necessary to fully answer this question.

Conclusion
This first prospective multicentre study demonstrates in routine clinical conditions, that [ 18 F]FDOPA PET adds information that can modify the management of high-grade glioma patients suspected of recurrence. When MRI findings are not straightforward, [ 18 F]FDOPA PET gives metabolic information complementary to MRI anatomical changes that contributes to better individual decisions during the course of the disease. These results are in favour of a wider use of [ 18 F]FDOPA PET in such situations.
All procedures performed in studies involving human participants were in accordance with the ethical standards of the institutional and/ or national research committee and with the 1964 Helsinki declaration and its later amendments or comparable ethical standards. Informed consent was obtained from all individual participants included in the study.