Maximum uptake and hypermetabolic volume of ¹⁸F-FDOPA PET estimate molecular status and overall survival in low-grade gliomas: a PET and MRI study

Abstract

Purpose:

We evaluated ¹⁸F-FDOPA positron emission tomography (PET) and magnetic resonance imaging (MRI) characteristics in association with the molecular status and overall survival (OS) in a large number of low-grade gliomas (LGGs).

Methods:

Eighty-six patients, who underwent ¹⁸F-FDOPA PET and MRI, and were diagnosed with new or recurrent LGGs were retrospectively evaluated with respect to their isocitrate dehydrogenase (IDH) and 1p19q status (10 IDH wild-type, 57 mutant, 19 unknown; 1p19q status in IDH mutant: 20 non-codeleted, 37 codeleted). After segmentation of the hyperintense area on fluid-attenuated inversion recovery image (FLAIR_ROI), the following were calculated: normalized maximum standardized uptake value (nSUV_max) of ¹⁸F-FDOPA relative to the striatum, ¹⁸F-FDOPA hypermetabolic volume (tumor-to-striatum ratios > 1), FLAIR_ROI volume, relative cerebral blood volume, and apparent diffusion coefficient within FLAIR_ROI. Receiver-operator characteristic and Cox regression analyses were performed.

Results:

PET and MRI metrics combined with age predicted the IDH mutation and 1p19q codeletion statuses with sensitivities of 73% and 76%, specificities of 100% and 94%, respectively. Significant correlations were found between OS and the IDH mutation status (hazard ratio [HR] = 4.939), nSUV_max (HR = 2.827), ¹⁸F-FDOPA hypermetabolic volume (HR = 1.048), and FLAIR_ROI volume (HR = 1.006). The nSUV_max (HR = 151.6) for newly diagnosed LGGs, and the ¹⁸F-FDOPA hypermetabolic volume (HR = 1.038) for recurrent LGGs demonstrated significant association with OS.

Conclusion:

Combining ¹⁸F-FDOPA PET and MRI with age proved useful for predicting the molecular status in patients with LGGs, while the nSUV_max and ¹⁸F-FDOPA hypermetabolic volume may be useful for prognostication.

Introduction

For clinical evaluation of low-grade gliomas (LGGs), magnetic resonance imaging (MRI) is the primary imaging modality owing to its high spatial resolution, high contrast within soft tissues, and no ionizing radiation. Numerous studies have reported the impact of MRI characteristics on tumor classification and prognosis; however, due to lack of contrast enhancement, LGGs cannot be fully evaluated by MRI alone. Meanwhile, radiolabeled amino acids positron emission tomography (PET), such as 3,4-dihydroxy-6-[18F]-fluoro-L-phenylalanine (¹⁸F-FDOPA), O-(2-[18F] fluoroethyl)-L-tyrosine (¹⁸F-FET), and [11C] methyl-L-methionine (¹¹C-MET), is often employed in neuro-oncological practice to identify metabolically active tissue.^{1 18}F-FDOPA and ¹⁸F-FET PET have improved distribution and efficiency owing to the relatively long half-lives of fluorinated tracers compared to carbon tracers.² They have been used for the evaluation of LGGs with regard to tumor classification,³ preoperative biopsy guidance,⁴ and prognostication.⁵ Because amino acids PET provides metabolic information to complement MRI, a combination of the techniques may yield more accurate observations for differentiating subtypes of LGGs and predicting prognosis than either technique alone.

Villani et al.⁵ reported that hyper ¹⁸F-FDOPA uptake was an independent predictor of progression-free survival; however, they did not evaluate the overall survival (OS). Patel et al.⁶ demonstrated that the combination of ¹⁸F-FDOPA and MRI characteristics is capable of predicting the degree of malignancy and OS among 45 gliomas, including 16 LGGs. The ability of ¹⁸F-FDOPA and MRI to predict OS in a large cohort of patients with LGGs has yet to be evaluated. In 2016, the World Health Organization (WHO) glioma classification was modified to include molecular subtypes including isocitrate dehydrogenase (IDH) gene mutations or chromosomal 1p19q codeletion.⁷ These molecular biomarkers have become essential for brain tumor classification, treatment decisions, and predicting prognosis for LGGs; however, the features of LGGs with different molecular subtypes are still debated.^{3, 8} Hence, there is a demand for non-invasive imaging biomarkers that can identify the molecular status and predict OS for LGGs.

The purpose of the current study was to evaluate characteristics of LGGs using ¹⁸F-FDOPA PET and multi-parametric MRI in a large patient cohort and determine associations between imaging metrics and their molecular status or OS.

Materials and methods

Patient Selection

Eighty-six patients with histologically confirmed LGGs with WHO grade II, who underwent ¹⁸F-FDOPA PET and MRI scans between 2007 and 2019, were retrospectively included. Selected MRI scans were performed within two months of the corresponding PET scans. MR perfusion imaging for 55 subjects and diffusion-weighted imaging (DWI) for 83 subjects, as well as conventional sequences, were obtained. All patients were diagnosed with new or recurrent LGGs according to the WHO 2007 or 2016 classification. When available, IDH1 mutational status, 1p19q codeletion status, and O6-methylguanine-DNA methyltransferase (MGMT) promoter methylation status were obtained.⁹ For newly diagnosed LGGs, no patients underwent stereotactic biopsy prior to ¹⁸F-FDOPA PET or MRI, and the median date between PET scan and surgery/biopsy was 18 days (range 1–505). OS was measured from the time of the PET scan until death or the censored date (median term 1272 days). This Health Insurance Portability and Accountability Act–compliant study has been approved by the institutional review board, and all subjects signed an informed consent form. The patient cohort was partly overlapped with a previous study.¹⁰

¹⁸F-FDOPA PET Image Acquisition

A ¹⁸F-FDOPA PET scan was performed with a full-ring PET/CT scanner (ECAT-HR; CTI/MIMVista) on the subjects, after they fasted for more than four hours. Following previously established procedures, ¹⁸F-FDOPA was synthesized and injected intravenously.¹¹ CT images were acquired prior to the PET scan for attenuation correction. Three-dimensional ¹⁸F-FDOPA emission data were acquired for a total of 30 minutes and integrated between 10 and 30 minutes following the injection to obtain 20-minute static ¹⁸F-FDOPA images. PET images were reconstructed using an ordered-subset expectation maximization iterative reconstruction algorithm, consisting of six iterations with eight subsets.¹² Finally, a Gaussian filter with a full width at half maximum of 4 mm was applied. The resulting voxel size was 1.34 × 1.34 × 3 mm for the ¹⁸F-FDOPA PET images. Standardized uptake value (SUV) maps of ¹⁸F-FDOPA were calculated based on the radioactive activity divided by the decay-corrected injected dose per body mass.¹³ The resulting SUV maps were subsequently normalized relative to the median value of the striatum (nSUV).¹⁴

MRI Acquisition

Anatomical MRI at least consisted of standard T1-weighted pre- and post-contrast images (2D axial turbo spin echo with 3-mm slice thickness and no interslice gap or 3D inversion prepared gradient echo images with 1–1.5 mm isotropic voxel size) and T2-weighted fluid-attenuated inversion recovery (FLAIR) images acquired at 3-mm slice thickness with no interslice gap on a 1.5-T or 3-T clinical MRI scanner.

For the dynamic susceptibility contrast perfusion MRI, a total dose of 0.1 mmol/kg of Gd-DTPA or Gd-BTDO3A (Magnevist or Gadavist; Bayer HealthCare Pharmaceuticals, Wayne, NJ) was administered. Relative cerebral blood volume (rCBV) maps were calculated using previously established procedures.¹⁵ Normalized rCBV maps were then computed by dividing the rCBV map by the median rCBV value of regions of interest (ROIs), placed on the contralateral normal-appearing white matter.

DWI was performed by a single-shot echo-planar imaging sequence in the axial plane with b = 1000 s/mm², slice thickness = 3 mm, and no interslice gap. Apparent diffusion coefficient (ADC) maps were calculated from the acquired images with b = 1000 s/mm² and b = 0 s/mm². In the case where DWI was not obtained, diffusion tensor imaging was used, acquired from 12–64 equidistant diffusion-sensitizing directions with b = 1000 s/mm² with a single b = 0 s/mm² image with slice thickness = 2–3 mm, and no interslice gap. Mean diffusivity maps were employed as estimates of the ADC using FSL software (dtifit; FMRIB, Oxford, UK; http://www.fmrib.ox.ac.uk/fsl/).

Postprocessing Analysis

All PET and MR images were registered to the corresponding post-contrast T1-weighted images using a six-degrees of freedom rigid transformation and a mutual information cost function using FSL (flirt). A single ROI as a FLAIR_ROI was semi-automatically segmented based on the regions of hyperintensity on the T2-weighted FLAIR images by a board-certificated neuroradiologist (H.T. with 13 years of clinical experience) using Analysis of Functional NeuroImages software (NIMH Scientific and Statistical Computing Core; Bethesda, MD; https://afni.nimh.nih.gov ).¹⁶ The maximum nSUV (nSUV_max) was quantified within the FLAIR_ROI. ¹⁸F-FDOPA hypermetabolic volume, including the voxel with nSUV > 1 within FLAIR_ROI, and FLAIR_ROI volume were calculated in milliliters. The median rCBV was calculated along with ADC low (ADC_low), which is defined as the lower mean of a double Gaussian mixed model fitted to the histogram of ADC values within the FLAIR_ROI.¹⁷

Figure 1 illustrates an example of segmentations of FLAIR_ROI and the ¹⁸F-FDOPA hypermetabolic region in a newly diagnosed 57-year-old male low-grade diffuse glioma patient.

Figure 1.

Postprocessing and segmentation example. A 57-year-old man with newly diagnosed astrocytic diffuse glioma (WHO grade II, IDH1 mutant, 1p19q non-codeleted, and MGMT methylated status). ROIs of the FLAIR hyperintense region (FLAIR_ROI, red area) and ¹⁸F-FDOPA hypermetabolic area (nSUV > 1, blue area) within FLAIR_ROI are shown. nSUV_max and volumes for each ROI are calculated. FLAIR_ROI is copied and pasted on rCBV and ADC maps, and median rCBV and ADC_low are calculated.

Statistical Analyses

The Shapiro-Wilk test was used to test for normality of the data. The student’s t-test for normally-distributed data and Mann-Whitney U test for non-normally-distributed data were performed.

Within the FLAIR_ROI, the evaluation included pairwise Spearman’s correlation between the nSUV_max, ¹⁸F-FDOPA hypermetabolic volume, FLAIR_ROI volume, median rCBV, and ADC_low, and the comparison of these values with regard to different molecular status (IDH mutation, 1p19q codeletion, and MGMT methylation status) and patient status (newly diagnosed and recurrent). The Kaplan-Meier curves and log-rank test were employed to compare OS for different molecular status. A multiple logistic regression model, integrating known clinical information such as age and MR-PET metrics, was used to predict the molecular status. Receiver operating characteristic (ROC) curves were used to determine whether a combination of clinical and MR-PET imaging information can discriminate between different molecular statuses. Area under the curve (AUC), along with the sensitivity and specificity of differentiation, was evaluated as a measure of model performance. Leave-one-out cross-validation was used to evaluate the accuracy of the multivariate logistic regression model.

Cox univariate regression analyses were conducted to investigate the association between OS and predictor variables including clinical information (sex, age, and molecular status) and imaging metrics (nSUV_max, ¹⁸F-FDOPA hypermetabolic volume, FLAIR_ROI volume, median rCBV, and ADC_low). For the Cox multivariate regression, the hazard of the nSUV_max, ¹⁸F-FDOPA hypermetabolic volume, and FLAIR_ROI volume controlling for age or molecular status (IDH or 1p19q) were evaluated separately because these three imaging variables were available from all subjects.

Additionally, subjects were stratified by molecular status and patient status, and the imaging metrics and their association with OS were evaluated for each subgroup analysis.

Statistical analysis was performed using R software (version 3.5.2; http://www.r-project.org/) and GraphPad Prism (version 8.3; GraphPad Software, La Jolla, CA). Statistical significance was defined as P < 0.05, and no correction for multiple comparisons was performed.

Results

Table 1 summarizes the patient demographics and molecular information, and Supplemental Table 1 describes this in more detail. The current study included 86 LGG patients (34 females) with a mean age of 43 years. Twenty-nine patients were newly diagnosed, while 57 patients had recurrent status. Ten gliomas were IDH wild-type (IDH_wt), 57 were IDH mutant (IDH_m), and 19 did not have confirmed IDH status. Among the IDH_m gliomas, 20 were 1p19q non-codeleted (IDH_m-non-codel), and 37 were 1p19q codeleted (IDH_m-codel).

Table 1.

Patient demographics and molecular information

No. of patients		86
Female		34 (39.5%)
Age ± standard deviation (year)		43.8 ± 12.6
IDH mutation and 1p19q codeletion status	IDH wild-type	10 (11.6%)
	IDH mutant 1p19q non-codeleted	20 (23.2%)
	IDH mutant 1p19q codeleted	37 (43.0%)
	Unknown	19 (22.1%)
MGMT-promoter methylation status	Unmethylated	17 (19.8%)
	Methylated	25 (29.1%)
	Unknown	44 (51.2%)
Patient status	Newly diagnosed	29 (33.7%)
	Recurrent	57 (66.3%)

IDH = isocitrate dehydrogenase; MGMT = O6-methylguanine-DNA methyltransferase

The pairwise Spearman’s correlation analysis (Figure 2) between the nSUV_max, ¹⁸F-FDOPA hypermetabolic volume, FLAIR_ROI volume, median rCBV, and ADC_low demonstrated a strong correlation between the nSUV_max and ¹⁸F-FDOPA hypermetabolic volume (r_s = 0.90), whereas the other pairs had weak or no correlations (−0.24 < r_s < 0.37).

Figure 2.

Pairwise Spearman’s correlation matrix between nSUV_max, ¹⁸F-FDOPA hypermetabolic volume, FLAIR_ROI volume, median rCBV, and ADC_low.

In the evaluation of patients stratified by the molecular status (Figure 3), the Kaplan-Meier curves and log-rank tests showed significant differences in the OS (P = 0.01). The IDH_m-codel subtype had the longest OS, followed by the IDH_m-non-codel subtype, and the IDH_wt subtype had the worst OS. When comparing the imaging metrics, the nSUV_max (P = 0.033) and ¹⁸F-FDOPA hypermetabolic volume (P = 0.043) were significantly higher in IDH_m-codel than IDH_m-non-codel. Other analyses did not yield any significant differences between the different molecular subtypes (all Ps > 0.06). The evaluation between MGMT unmethylated and methylated subtypes was shown in Supplemental Figure 1.

Figure 3.

a) Kaplan-Meier plots showing significant differences in OS between IDH_wt, IDH_m-non-codel, and IDH_m-codel gliomas, or between IDH_wt and IDH_m. b) The nSUV_max and c) volume of ¹⁸F-FDOPA hypermetabolic area and FLAIR_ROI, and d) median rCBV and ADC_low between different molecular statuses. The nSUV_max and ¹⁸F-FDOPA hypermetabolic volume are significantly higher in IDH_m-codel than IDH_m-non-codel. When comparing between IDH_wt and IDH_m (including IDH_m-non-codel and IDH_m-codel), no imaging metrics show significant differences. ROC curves for differentiating e) IDH mutation or f) 1p19q codeletion status using patient age and imaging information, only age, both PET and MRI, only PET, or only MRI.

Subsequently, we tested whether the combination of the MR and PET imaging metrics and patient age can be used to predict the molecular status. We created a new metric from a combination of the imaging factors and patient age by incorporating a multiple logistic regression model as follows:

$$\begin{gathered} \ln \left[\frac{P(wt)}{P(m)}\right]=0.295\cdot (Age)-0.820\cdot \left(nSU{V}_{max}\right)-0.241\cdot \text{(}FDOPAHypermetabolicVolume[mL]) \\ +1.026\times {10}^{-2}\cdot \left(FLAI{R}_{ROI}Volume[mL]\right)+0.966\cdot (rCBV) \\ +1.783\times {10}^{-2}\cdot \left(AD{C}_{low}\left[\times {10}^{-6}\right]\right)-\underset{Intercept}{\underbrace{34.90}} \end{gathered}$$

This metric enabled to differentiate the IDH mutation status (AUC, 0.93, sensitivity, 73%, specificity, 100%), with only patient age as a significant factor (odds ratio = 1.34, P = 0.04). Leave-one-out classification accuracy to differentiate IDH mutation was 86% (sensitivity, 97%, specificity, 71%). Similarly, for prediction of the 1p19q codeletion status, a multiple logistic regression analysis was performed as follows:

$$\begin{gathered} \ln \left[\frac{P(codel)}{P(noncodel)}\right]=0.237\cdot (Age)+1.360\cdot \left(nSU{V}_{max}\right)+0.148\cdot \text{(}FDOPAHypermetabolicVolume[mL]) \\ \begin{array}{l}-0.760\cdot \left(FLAI{R}_{RoI}Volume[mL]\right)-0.721\cdot (rCBV)\hfill \\ -1.102\times {10}^{-3}\cdot \left(AD{C}_{low}\left[\times {10}^{-6}\right]\right)-\underset{Intercept}{\underbrace{7.111}}\hfill \end{array} \end{gathered}$$

This metric enabled to differentiate the 1p19q codeletion status (AUC, 0.91, sensitivity, 76%, specificity, 94%), with patient age (odds ratio = 1.27, P = 0.011) and FLAIR_ROI volume (odds ratio = 0.93, P = 0.019) being significant factors. Leave-one-out classification accuracy to differentiate 1p19q codeletion was 65% (sensitivity, 63%, specificity, 67%). ROC curves to predict IDH or 1p19q status were also evaluated using only the patient age, only PET metrics (i.e., nSUV_max and ¹⁸F-FDOPA hypermetabolic volume), only MRI metrics (i.e., FLAIR_ROI volume, median rCBV, and ADC_low), or combined PET and MRI metrics. The AUC of IDH or 1p19q status was higher using both PET and MRI parameters than when using either of the parameters individually; however, the AUC incorporating patient age and MR-PET parameters yielded the highest value.

When stratifying the patients into newly diagnosed and recurrent LGGs (Figure 4), the nSUV_max, ¹⁸F-FDOPA hypermetabolic volume, and ADC_low were significantly higher in the recurrent LGGs than newly diagnosed LGGs (P < 0.001, < 0.001, and 0.016, respectively). Further subgroup analyses, stratified by molecular status (IDH_wt, IDH_m-non-codel, and IDH_m-codel) and patient status (newly diagnosed and recurrent), are shown in Supplemental Figure 2, and exhibited similar trends with higher nSUV_max and ¹⁸F-FDOPA hypermetabolic volume in IDH_m-codel than IDH_m-non-codel in both newly diagnosed and recurrent groups, although they were not significant.

Figure 4.

a) The nSUV_max, b) volume of ¹⁸F-FDOPA hypermetabolic area or FLAIR_ROI, and c) median rCBV and ADC_low between newly diagnosed and recurrent LGGs.

The Cox univariate analysis (Table 2) showed a significant increase in hazard associated with the IDH status (hazard ratio [HR] = 4.939, 95% confidence interval [CI] = [1.203–20.28], P = 0.026), nSUV_max (HR = 2.827, CI = [1.202–6.649], P = 0.017), ¹⁸F-FDOPA hypermetabolic volume (HR = 1.048, CI = [1.019–1.078], P = 0.001), and FLAIR_ROI volume (HR = 1.006, CI = [1.000–1.013], P = 0.046). The Cox multivariate regression analysis controlling for age showed a significant increase in the hazard associated with the nSUV_max (HR = 3.208, CI = [1.272–8.090], P = 0.013) and ¹⁸F-FDOPA hypermetabolic volume (HR = 1.045, CI = [1.017–1.073], P = 0.001), but not with FLAIR volume (HR = 1.006, CI = [0.999–1.013], P = 0.08). When controlling for IDH status or 1p19q status (for IDH_m LGGs), the Cox multivariate regression analysis showed a significant increase in the hazard of OS associated with the nSUV_max (controlling for IDH status: HR = 4.101, CI = [1.244–13.53], P = 0.020; controlling for 1p19q status: HR = 6.169, CI = [1.537–24.76], P = 0.010) and ¹⁸F-FDOPA hypermetabolic volume (controlling for IDH status: HR = 1.063, CI = [1.005–1.125], P = 0.032; controlling for 1p19q status: HR = 1.097, CI = [1.021–1.180], P = 0.011), but not with FLAIR volume (controlling for IDH status: HR = 1.002, CI = [0.981–1.024], P = 0.83; controlling for 1p19q: HR = 1.003, CI = [0.981–1.025], P = 0.80). Subgroup analysis stratified by different molecular status or patient status demonstrated that the nSUV_max represented an independent predictor of OS for IDH_m-non-codel (HR = 6.100, CI = [1.155–32.22], P = 0.033) and newly diagnosed LGGs (HR = 151.6, CI = [1.289–17830], P = 0.038), while the increase in the ¹⁸F-FDOPA hypermetabolic volumes (HR = 1.038, CI = [1.009–1.067], P = 0.009) and FLAIR_ROI volumes (HR = 1.006, CI = [1.000–1.013], P = 0.049) were independent predictors for recurrent LGGs (Supplemental Table 2).

Table 2.

Cox univariate regression

	Cox univariate regression
	Hazard ratio	95% confidence interval	P value
Sex	1.579	0.551–4.517	0.39
Age (year)	1.03	0.996–1.064	0.083
IDH status (wild-type vs. mutant)	4.939	1.203–20.28	0.026*
1p19q status for IDH mutant (non-codel vs. codel)	0.234	0.042–1.301	0.097
MGMT Status (unmethylated vs. methylated)	0.56	0.078–4.001	0.56
Patient Status (newly diagnosed vs. recurrent)	2.231	0.640–7.766	0.2
nSUVmax	2.827	1.202–6.649	0.017*
FDOPA hypermetabolic volume (mL)	1.048	1.019–1.078	0.001*
FLAIRROI volume (mL)	1.006	1.000–1.013	0.046*
rCBV	1.05	0.214–5.151	0.95
ADClow (×10−6 mm2/s)	1.002	0.999–1.005	0.17

IDH = isocitrate dehydrogenase; MGMT = O6-methylguanine-DNA methyltransferase; rCBV = relative cerebral blood volume; ADC = apparent diffusion coefficient

^*means statistically significant

Discussion

In the current study, ¹⁸F-FDOPA PET and MRI characteristics were evaluated with regard to their association with molecular status and OS in LGG patients. Our investigation revealed that combining PET and MRI information with patient age could predict IDH and 1p19q status more accurately than using PET or MRI information alone. We also confirmed that nSUV_max and ¹⁸F-FDOPA hypermetabolic volume were higher in IDH_m-codel than IDH_m-non-codel LGGs. The Cox regression analysis revealed significant associations of OS with the IDH status, nSUV_max, ¹⁸F-FDOPA hypermetabolic volume, and FLAIR_ROI volume. Because not all recurrent glioma patients in all institutions have a known molecular status at the time point of the previous surgery or biopsy, the results of predicting molecular status and prognosis are useful not only for newly diagnosed LGGs, but also recurrent LGGs.

The molecular status of LGGs is crucial for patient diagnosis and predicting prognosis. As shown previously and in this study, the IDH mutation status was significantly associated with OS with shorter survival for IDH_wt LGGs due to the biological similarities to glioblastomas.^{8, 18} In the evaluation of imaging metrics of LGGs, although some studies have evaluated the association of amino acid tracer uptake and molecular status, the imaging features in different molecular status were not confirmed. One study stratified gliomas into IDH_wt and IDH_m groups, which included 1p19q non-codeleted and codeleted gliomas, and showed higher tracer uptake in IDH_m than IDH_wt.³ Other studies stratified gliomas into 1p19q codeleted and non-codeleted groups, which included both IDH_m-non-codel and IDH_wt gliomas, and showed no significant differences between 1p19q codeleted and non-codeleted gliomas.^{19, 20} The diverse patient cohort in each study may confuse the issue and mask the associations of the amino acid tracer uptake with molecular status. In contrast, the current study stratified gliomas into three groups (IDH_wt, IDH_m-non-codel, and IDH_m-codel) and confirmed the features of amino acid tracer uptake, including higher nSUV_max and ¹⁸F-FDOPA hypermetabolic volume in IDH_m-codel than IDH_m-non-codel LGGs. This was consistent with previous literatures indicating a higher tracer uptake in oligodendrogliomas compared with astrocytomas, reflecting higher cell density, endothelial hyperplasia, microvascular proliferation, and higher vascular bed in oligodendroglial compared to astrocytic components.^{8, 10, 21, 22} When comparing between IDH_wt and IDH_m, the ratio of IDH_m-codel LGGs in the IDH_m group may affect the differences of ¹⁸F-FDOPA uptake because the current study revealed IDH_m-codel LGGs to show high tracer uptake.

In the current study, the combination of PET and MRI information with patient age successfully differentiated the IDH mutation and 1p19q codeletion status with AUCs higher than 0.90, although the performance, particularly for the IDH mutation status, was largely dependent on the patient age. A previous study revealed that the IDH_m-non-codel group tended to be found in the youngest patients, followed by IDH_m-codel and IDH_wt;²³ hence, the combination of PET and MRI, along with patient age may be helpful for predicting the molecular status.

High ¹⁸F-FDOPA uptakes are known to reflect high metabolic activity and predict worse survival outcomes in both newly diagnosed and recurrent gliomas.^{6, 24} For glioblastomas, hypermetabolic volume was an important factor.²⁵ For newly diagnosed LGGs, the uptake of amino acid tracers was reported to be associated with progression-free survival^{5, 8} and predicted disease progression after 1-year follow-up.²⁶ No studies have revealed the association of ¹⁸F-FDOPA uptake and OS in grade II gliomas alone, partly due to a small cohort population. The current large cohort study revealed that increased nSUV_max of ¹⁸F-FDOPA was associated with a worse prognosis in grade II gliomas; furthermore, the ¹⁸F-FDOPA hypermetabolic volume was also associated with a worse prognosis. These results were consistent with the strong correlation between the nSUV_max and ¹⁸F-FDOPA hypermetabolic volume, detected in this study. Previous studies have indicated that volume of the contrast-enhancing regions was predictive of the OS;²⁷ however, due to the lack of contrast enhancement, such regions cannot be evaluated in most grade II gliomas. Meanwhile, because hypermetabolic volume can be calculated regardless of contrast enhancement, it could be a useful biomarker for gliomas, especially for the non-contrast-enhancing LGGs.

When comparing the newly diagnosed and recurrent LGGs, both nSUV_max and ¹⁸F-FDOPA hypermetabolic volume were significantly higher in recurrent LGGs. The treatment-related changes may have affected amino acid tracer uptake, because ¹⁸F-FDOPA uptake in normal-appearing brain structures might be altered by temozolomide treatment.²⁸ The blood-brain barrier breakdown, due to cancer progression, may contribute to the extent of amino acid transport in recurrent gliomas, suggesting that the SUV may not directly reflect recurrent tumor activity.²⁹ Malignant transformation at recurrence may also induce an increase in the SUV in some LGGs.²⁴ Unfortunately, the histopathology of recurrent tumors was not available for all patients, and thus it was not analyzed in this study. Several studies evaluated the longitudinal change of amino acids tracer uptake, reporting that a higher rate of temporal change in the ¹⁸F-FDOPA uptake was associated with a higher risk of malignant transformation and poor survival in patients with LGGs.⁹ A study reported that 65% of primary gliomas with a negative ¹⁸F-FET uptake, which could not be delineated from the background brain tissue, turned PET-positive during follow-up scans, indicating that gliomas can change their ¹⁸F-FET uptake behavior throughout the course of the disease.³⁰ Although ¹⁸F-FET PET-negative gliomas in general have a better prognosis than ¹⁸F-FET positive gliomas, ¹⁸F-FET PET-negative gliomas with photopenic defects were reported to have a higher risk of harboring a higher-grade glioma and an unfavorable outcome than gliomas with indifferent ¹⁸F-FET uptake to the background.^{31, 32} Meanwhile, this study exhibited a better prognosis in patients with lower ¹⁸F-FDOPA uptake, and no studies have reported such unfavorable outcomes for ¹⁸F-FDOPA hypo-uptake gliomas. The different disease courses may be caused by different amino acid tracers, in particular different metabolic processes of ¹⁸F-FDOPA and ¹⁸F-FET. Hence, comparison of the glioma pathology and amino acid uptakes with different tracers is desirable, in a longitudinal manner.

The retrospective nature of this study presents one of its limitations; specifically, the clinical information (Karnofsky Performance Status), molecular status, and rCBV/ADC maps were not obtained for all subjects, and the imaging protocols were not exactly matched. Although examination and treatment planning were discussed in weekly tumor boards at our institution, the patient cohort was potentially influenced by selection bias because FDOPA PET examination may have been performed more often for glioma patients who were suspected to have primary or recurrent gliomas but were difficult to be diagnosed based on conventional MRI alone. For three patients with newly diagnosed LGGs, the interval between PET and surgery/biopsy was longer than half a year; however, they did not receive additional treatments between the interval, and the WHO grades remained at grade II when the pathology was confirmed after surgery. Nonetheless, the possibility of temporal change in molecular status during the interval cannot be excluded. Including patients with various previous treatment statuses may have influenced the MRI and/or PET imaging features. As we could not obtain pathologies of all recurrent gliomas after the PET examination, the WHO grade may have been underestimated. Although this study used leave-one-out cross validation for evaluating the predictive performance of molecular status, another independent cohort is required to generalize our classification performance.

Conclusion

This was the largest population study to date evaluating LGGs using both ¹⁸F-FDOPA PET and MRI. A combination of PET, MRI, and patient age may be helpful for predicting the molecular status in patients with LGGs, and ¹⁸F-FDOPA PET metrics proved useful for estimating the OS.