Patients
From October 2012 to October 2017, 245 patients referred to the Department of Nuclear Medicine at the CHRU Nancy for brain tumor assessment were investigated by 18F-FDopa PET. Fifty-one of these patients were retrospectively selected on the basis that they had an initial history of surgically-resected glioma and that the considered 18F-FDopa PET had been prescribed for differentiating recurrence/tumor progression from post-therapeutic effects after a non-contributive MRI. As usual in our department, a minimum delay of 2 months is always respected between the surgery or end of radiotherapy and the performing of 18F-FDopa PET in order to reduce the risk of 18F-FDopa PET false positives [18].
In such patients, a clinical follow-up and MRI are systematically performed at least every 3 months or at shorter intervals as clinically indicated after surgery [8]. For the present study, the final diagnosis of glioma recurrence/progression at 6 months or for further assessment of survival was blinded from the 18F-FDopa PET results and was based on current guidelines for which a recurrence/progression is the result of any new tumor or brain lesion at MRI and/or clear increase in tumor size or in contrast enhancement, and/or significant clinical deterioration, with all of these criteria not being attributable to other non-tumor causes and not due to steroid tapering [19–21]. Patients with lesions which were classified as therapy-related changes due to the course of the disease did not receive any additional therapy beyond the standard therapy due to suspected recurrence. For 4 patients having undergone stereotactic biopsy or open resection, the diagnosis of recurrence/prognosis was assessed neuropathologically.
The assessment of 18F-FDopa PET parameters for differentiating recurrence/progression from post-therapeutic effects was based on the evaluation of the previous criteria during a 6-month follow-up period. However, PFS and overall survival (OS) were calculated from the date of the PET exam to the date of definite diagnosis of progression and of death, respectively, with a minimum delay of 19 months of observation. The final date for reporting any event for PFS or OS was June 1st, 2019.
The local ethics committee (Comité d’Ethique du CHRU de Nancy) approved the retrospective data evaluation on June 7, 2018, and authorization from the CNIL (National Commission on Information Technology and Civil Liberties) was delivered on June 25, 2018 (authorization n° R2018-11). This study complied with the principles of the Declaration of Helsinki. Informed consent was obtained from all individuals included in the study.
Initial pathological grading of the gliomas
All cases were reviewed and classified according to the WHO 2016 classification from tumor samples provided by surgery or stereotactic biopsy [22]. IDH mutation status was assessed by immunohistochemistry with IDH1 R132H protein expression (Dianova, clone H09), or by Sanger sequencing in case of ATRX immunohistochemical loss without IDH1 R132H staining [23]. Tumors presenting oligodendroglial morphology or showing IDH mutation without ATRX loss were additionally tested for 1p/19qco-deletion using multiplex PCR fragment analysis (loss of heterozygosity), or comparative genomic hybridization [24].
PET recordings and image reconstruction
18F-FDopa PET-computed tomography (CT) scans were obtained on a Biograph hybrid system involving a six-detector CT for attenuation correction (Biograph 6 True Point, SIEMENS, Erlangen, Germany). Patients were instructed to fast for at least 4 hours with some patients also receiving Carbidopa administration one hour prior to their exam (n = 17). The CT scan was recorded first and immediately followed by a 30-min 3D list-mode PET recording initiated during the bolus injection of 3 MBq of 18F-FDopa per kilogram of body weight. The static PET images were reconstructed from the list mode data ranging from 10 to 30 min post-injection [4,7] while the PET images reconstructed for dynamic parameters encompassed 6 consecutive frames of 20 sec each followed by 28 frames of 1 min each. The choice of this acquisition time frame was based on previous studies performed with 18F-FET PET (from 0 to 40 minutes post-injection with a reconstructed 20- to 40-min static image [25]) and on the maximum observed uptake of 18F-FDOPA in a PET study involving high-grade and low-grade gliomas (respectively 8 and 10 minutes post-injection) [26].
All static and dynamic images were reconstructed with an OSEM 2D algorithm (2 iterations, 21 subsets, 4-mm Gaussian post-reconstruction filter), corrected for attenuation, scatter and radioactive decay, and displayed in a 256x256 matrix with 2.7x2.7x3.0 mm3 voxels.
Analyses of PET images
Regions of interest (ROIs) were placed on the static PET images using a dedicated software (DOSIsoft, Cachan, France). A spherical ROI of 2 cm diameter, centered on the maximum lesion uptake, was used for determining maximum and mean Standardized Uptake Values (SUVmax and SUVmean, respectively). Tumor-to-striatum (TSR) and tumor-to-normal-brain (TBR) ratios were computed as SUVmean or SUVmax of the lesion uptake divided by the SUVmean of the striatum (TSRmean and TSRmax) or of normal brain (TBRmean and TBRmax). The SUVmean from the striatum was obtained from the contralateral basal ganglia, which was segmented using a threshold of 65% of maximal uptake, while the normal reference brain SUVmean was obtained with a crescent shape ROI (2x8 cm) positioned on the semi oval center of the unaffected contralateral hemisphere, including white and grey matter [4].
When no abnormal 18F-FDopa uptake was detected, the ROIs of the potential residual tumor were placed at the site of maximal MRI abnormalities with a fused display of PET and Fluid Attenuation Inversion Recovery (FLAIR) MRI images [7].
As previously described [27], the metabolic tumor volume (MTV) was obtained through a 3D auto-contouring process with a threshold corresponding to the SUVmean of the contralateral striatum.
In addition, time-activity curves, representing the evolution of the TBRmean as a function of time (TACratio), were extracted with the PLANET® Dose software (DOSIsoft, Cachan, France) and with the ROIs previously placed on static images (see above). Each dynamic frame was previously registered on the CT images, in order to take into account potential patient movements during acquisitions [28]. Two dynamic parameters were determined from fitted curves to overcome noise effects, using a method already validated for 18F-FET in the same setting [29], namely: i) Time-To-Peak (TTP), corresponding to the delay between the onset of the dynamic acquisition (time of tracer injection) and the time-point of the maximal TBRmean value, and ii) slope, calculated with a linear regression applied from the 10th to 30th minute.
Statistical analysis
Categorical variables are expressed as percentages and continuous variables as median and interquartile range due to the non-normality of variable distributions.
Recurrence/progression at 6 months follow-up
Univariate analysis was performed using Mann-Whitney tests applied between patients with glioma recurrence/progression and the remaining patients. In order to calculate diagnostic performances, the optimal threshold for each static and dynamic PET parameter was determined from ROC curves using the maximal value of the product of sensitivity and specificity. Thereafter, a multivariable logistic regression model with forward selection was performed for predicting patients with glioma recurrence/progression.
Progression-free survival (PFS) and Overall survival (OS)
The dichotomized parameters, which were determined according to previously mentioned optimal thresholds, were used in survival analyses. For this purpose, univariate survival probabilities according to the Kaplan-Meier method with the log-rank test used and the hazard ratio interval of each parameter with its 95% confidence interval were calculated to compare survival curves.
P-values obtained in univariate analysis as well as in survival analysis were adjusted using Benjamini-Hochberg correction in order to reduce the risk of false discovery [30]. P-values lower than 0.05 were considered as significant.
Analyses were performed with SPSS (SPSS Statistics for Windows, Version 20.0. Armonk, NY: IBM Corp) and R (R Foundation for Statistical Computing, Vienna, Austria).