Study population
In this retrospective study, from November 2014 to April 2018, a total of 338 patients were referred to undergo 11C-MET PET scans for “brainstem space-occupying lesions” as an initial diagnosis. We referred to the database of the National Brain Tumour Registry of China (NBTRC) established by the surgeon employed in our hospital and selected patients who had detailed survival information and were diagnosed with DIPG. DIPG was defined as a T1 hypo- (or iso-) intense and T2-hyperintense tumour involving at least 50% of the pons on MRI. Exclusion criteria included (1) aged older than 18 years; (2) post-treatment (surgery, radiotherapy, or systemic therapy) PET; and (3) poor imaging quality. This study was approved by the Ethics Committee of the Beijing Tiantan Hospital, Capital Medical University, and the requirement for informed consent was waived by the Institutional Reviewing Board of Beijing Tiantan Hospital, Capital Medical University.
Basic patient information, tumour histology (if available), and therapeutic information were also obtained from the NBTRC database. Basic patient information included: sex, age, symptom duration, and Karnofsky performance status (KPS). Tumour grading was determined according to the revised 2016 WHO criteria for this study was conducted before the new WHO classification published in 2021 [28, 29]. Treatment consisted of systemic therapy and radiotherapy. In particular, the H3 K27M mutation status (if available) was recorded for further analysis.
Imaging protocol and analysis
A total of 555–740 MBq of 11C-MET was intravenously injected, and whole-brain imaging in the 3-dimensional mode was started 10 min later. 11C-MET PET images were acquired for 10–15 min using a PET/CT scanner (Elite Discovery; GE Healthcare, Milwaukee, WI, USA) with a 5-mm axial resolution and full-width-at-half-maximum at the centre of the field of view of 4 mm. A nondiagnostic low-dose CT scan was used for attenuation correction. Imaging data were reconstructed into 30 axial planes with a slice thickness of 5 mm and a 192x192 image matrix.
11C-MET PET images were windowed to visualise focal 11C-MET avidity by a board-certified nuclear medicine physician. The uptake intensity of 11C-MET in tumours was graded on a scale from 1 to 3, with 1, 2, and 3 indicating that uptake was less, approximately equal, and greater than that of the background, respectively. Uninvolved left frontal background white matter was defined as the background [22]. Further, uptake intensity with a scale of 3 and 1–2 were considered PET-positive and negative uptake, respectively.
11C-MET uniformity was defined as the percentage of tumour volume exhibiting 11C-MET uptake and was graded on a 4-point scale (1, 1%– 24%; 2, 25%– 49%; 3, 50%– 74%; 4, 75%– 100%) [22]. Tumour volume was delineated on MRI fluid-attenuated inversion recovery (FLAIR) images. The uniformity was determined using visual estimation. For cases with PET-negative uptake in tumours, the uniformity was considered to be grade 1. Furthermore, a scale of 1–3 was deemed low uniformity grade, while a scale of 4 was considered high uniformity grade.
Standardised uptake values (SUV), metabolic tumour volume (MTV, expressed in cubic centimetres; cc), and total lesion methionine uptake (TLMU, expressed in cubic centimetres; cc) were measured using PETVCAR, which is an automated segmentation software system that uses an Advantage Workstation (version 4.6; GE Healthcare, Milwaukee, WI, USA). Threshold-based volume of interest (VOI) analysis was performed for all 11C-MET PET images. A threshold of 1.3 times the mean reference tissue activity was used to delineated the tumour extent. The mean reference tissue activity was measured from a large crescent-shaped VOI, which was drawn in the normal cerebral hemisphere at the level of the centrum semiovale, including the cortical and white matter [30]. A VOI around the whole tumour was auto-contoured and segmented using a boundary box, which was placed by a board-certified nuclear medicine physician who adjusted to ensure that this 3-dimensional cube contained the entire tumour. If this automated contouring procedure produced a tumour area that contained a significant amount of the normal brain tissue (such as a developmental venous anomaly, the pituitary gland, and the basilar clivus), the nuclear medicine physician referred to the MRI FLAIR images and performed manual correction in the axial, sagittal, and coronal planes.
The maximum of SUV (SUVmax), mean SUV (SUVmean), peak SUV (SUVpeak), MTV and TLMU in the VOI of the tumours were recorded. For PET-negative patients, we also calculated SUV, MTV, and TLMU using the above-mentioned threshold-based VOI analysis referring to MRI FLAIR images, and the data were included in the analyses. The tumour-to-background ratio (TBR) was defined as the SUVmax, SUVmean, and SUVpeak of tumours divided by the mean reference tissue activity.
MRI images were obtained using GE 3.0T scanners (Genesis Signa and Signa HDe) and Siemens 3.0T scanners (Trio Tim and Verio). Each patient underwent routine clinical MRI scans including pre-contrast T1-weighted, FLAIR, and T2-weighted images. Following gadolinium compound bolus administration (0.1 mmol/kg, macrocyclic ionic agent), axial, coronal, and sagittal SE T1-weighted images were acquired. The MRI contrast enhancement patterns of the lesions were evaluated as present or absent by an experienced neuroradiologist (A.l.).
Statistical analysis
Descriptive statistics included the median, minimum, and maximum continuous parameters. For categorical parameters, the number and percentage distributions were used.
Due to the non-normality of the data, Mann–Whitney U tests were used to evaluate the association of H3 K27M mutation and H3 K27M WT changes among PET continuous variables. The χ2 test or Fisher’s exact test was used to evaluate the association of H3 K27M mutation and H3 K27M WT changes among PET and MRI categorical variables.
PFS was calculated from the date of diagnosis to the date of disease progression. Progression was determined according to the recommendation of the Response Assessment in Paediatric Neuro-Oncology (RAPNO) Working Group [31]. OS was defined as the interval between the date of diagnosis and the date of the last follow-up or death. Patients who experienced no event were censored at the last follow-up date.
Probability estimates of PFS and OS were calculated using the Kaplan–Meier method and compared using the log-rank test. Cox proportional hazards regression was used to identify imaging and clinicopathological predictors of PFS and OS distributions after adjusting for sex and age. The results are summarised as PFS and OS with a hazard ratio (HR). Only the covariates that were significant at the p༜0.05 level in univariate analysis were entered into multivariate analysis. The proportional hazards assumption was verified graphically. Two-tailed probabilities were reported, and a p value of 0.05 was used to define nominal statistical significance; given the exploratory nature of the study, no multiple testing corrections were applied.
Correlations between the PET parameters (TBRs, MTV, and TLMU) were assessed using Pearson’s correlation analysis. For the highly correlated PET parameters, the parameter with the smallest univariate p- value was selected for the multivariate model. Statistical analyses were performed using R software (version 3.1.3) and SPSS 21.0.