Performance of ¹⁸F-FET-PET versus ¹⁸F-FDG-PET for the diagnosis and grading of brain tumors: inherent bias in meta-analysis not revealed by quality metrics

We read with great interest the recent paper by Dunet et al.¹ The authors performed a meta-analysis incorporating data from 5 studies (119 patients) to assess the performance of ¹⁸F-FET-PET versus ¹⁸F-FDG-PET for the diagnosis and grading of brain tumors. The authors concluded that ¹⁸F-FET-PET performed much better than ¹⁸F-FDG-PET and should be used when assessing a new brain tumor.

We commend the authors for performing a meta-analysis on such an important topic since ¹⁸F-FET-PET has shown promising results in the diagnosis of brain tumors.² There is a limited amount of data in the literature comparing ¹⁸F-FET-PET directly with ¹⁸F-FDG-PET; however, we have several concerns. Although study quality (assessed with STARD and QUADAS criteria) and all of the metrics for measuring bias seem reasonable, there is a considerable amount of heterogeneity in the indications for imaging, cohort characteristics, and methodologies among the 5 studies included in the meta-analysis (Table 1).

Table 1.

Five studies included in the meta-analysis comparing 18F-FET-PET and 18F-FDG-PET

Study	Cohort Size	Both FET and FDGa	Indication for Imaging	Interval between FET and FDG (days)	Tumorb	Gliomac	High Graded	Pathological Correlation
Lau et al	21	21	Suspected or known brain tumor	Separate days with a mean of 4 days	71% (15/21)	52% (11/21)	27% (3/11)	57% (12/21)
Pauleit et al	52	52	Suspicion of a cerebral glioma or recurrence of a previously operated glioma	Together	87% (45/52)	83% (43/52)	49% (21/43)	100% (52/52)
Floeth et al	14	11	Suspicion of a cerebral glioma or recurrence of a previously operated glioma	Different day within a week	36% (5/14)	36% (5/14)	100% (5/5)	93% (13/14)
Plotkin et al	15	15	Non-contrast–enhancing gliomas	<4 week (<1 week in 8/15)	100% (15/15)	100% (15/15)	40% (6/15)	93% (13/14)
Pichler et al	88	23	New lesion on MRI	Not stated	98% (59/60)	85% (51/60)	67% (34/51)	68% (60/88)

^aPatients who underwent both ¹⁸F-FET-PET and ¹⁸F-FDG-PET.

^bTumor: the percentage of tumor on final histopathology.

^cGlioma: the percentage of glioma on final histopathology.

^dHigh grade: the percentage of high-grade glioma on final histopathology.

First, the indication for imaging was significantly different among the studies included in the meta-analysis. For example, Floeth et al included patients with solitary intracerebral lesions showing ring enhancement on contrast-enhanced MRI in order to distinguish a malignant lesion (eg, glioblastoma) from a benign one (eg, abscess).³ This resulted in a low percentage of tumors/gliomas (36%) on histopathology compared with other studies. In contrast, in the study by Plotkin et al, only patients with gliomas or lesions highly suspicious of a glioma recurrence were included.⁴ As a result, the percentage of tumors/gliomas was 100%. In addition, Plotkin et al only included lesions that were non-contrast–enhancing on MRI. In the studies by Lau et al and Pauleit et al, the indication for imaging was suspected or known brain tumor.^5,6 However, in the study by Pichler et al, patients with known history of brain tumor or cranial surgery of any kind were excluded.⁷ Thus, the pretest probability of a lesion being diagnosed by ¹⁸F-FET-PET or ¹⁸F-FDG-PET as a brain tumor varied from study to study due to differences in the cohort characteristics.

Second, there were significant differences in the imaging protocols and analyses used in the 5 studies. For example, in the study by Pauleit et al, ¹⁸F-FET-PET and ¹⁸F-FDG-PET were performed on the same day, and the cerebral accumulation of 18F-FDG-PET was calculated by decay-corrected subtraction of the ¹⁸F-FET-PET scan from ¹⁸F-FET-PET/¹⁸F-FDG-PET scan.⁶ In the other 4 studies, ¹⁸F-FET-PET and ¹⁸F-FDG-PET were performed on separate days. In addition, the region of interest (ROI) was chosen and registered differently between ¹⁸F-FET-PET and MRI among the studies, which would affect the calculation of tumor-to-background ratio (TBR). For example, in the study by Pauleit et al, the ROI for the contralateral brain was placed on the white matter only for ¹⁸F-FDG-PET, since (as claimed by the authors) ¹⁸F-FDG-PET uptake in the brain differs considerably between white and gray matter.⁶ However, in the study by Floeth et al, reference ROI was placed on contralateral normal-appearing gray matter for ¹⁸F-FDG-PET.³ In the other 3 studies, whether the control ROI was placed on the gray or white matter was not specified. The variations in imaging protocols and how the ROI locations were determined certainly affected the quantitative analyses of TBR in which individual values from different studies that used different methods were pooled together.

In conclusion, despite the seemingly acceptable quality metrics and bias analyses for the meta-analysis, variations in patient selection and imaging protocol/analyses significantly limit the validity of the conclusions derived from its results.

Funding

This work was supported by the Natural Science Foundation of China (81301988 to Li Yang), China Ministry of Education Doctoral Program Spot Foundation (20130162120061 to Li Yang), and the Shenghua Yuying Project of Central South University, 2015 (to Li Yang).

Conflict of interest statement. The authors declare no conflicts of interest.