Imaging practice in low-grade gliomas among European specialized centers and proposal for a minimum core of imaging

Ellingson et al. [1] proposed a standardized brain imaging protocol for tumor patients. We support the principle of creating standardized protocols for DLGG patients. This should include minimum imaging datasets, recommended imaging frequency and recommendations about additional sequences. The goal of establishing a minimal core of imaging would be to allow further investigations with uniform imaging throughout different countries. This would enable to include numerous patients in future prospective cohorts, ensuring a sufficient volume of data to perform big data analysis [14].

The minimal core of imaging in DLGG needs to take regional and national differences regarding technical standards and reimbursement into account. Therefore, the following imaging algorithm is proposed: MR imaging should incorporate sequences for morphologic descriptive analysis and those focusing on potential malignant transformation. T1wo/w and T2 images were used by all centers and therefore represent the cornerstone of morphologic imaging. In addition, TIRM or FLAIR sequences should be obtained and can also be used for volumetric assessment of DLGG. Although several recommendations set the minimum level of slice thickness in T2 and TIRM/FLAIR imaging to ≤ 4 mm [1], it is known that slice thickness is key for accurate volumetric and treatment response assessment [15‐17]. Perfusion weighted imaging (PWI) is used frequently for assessment of treatment response and detection of anaplastic transformation [18, 19]. In our survey, response and transformation were predominantly determined with MRI. T1 (with contrast) and PWI were used in 87 and 55%, respectively. Diffusion weighted imaging (DWI), and especially the computed ADC has been used to investigate malignant transformation [20] and is applied in up to 90% of centers.

We therefore recommend that the minimal MRI sequence dataset should consist of T1wo/w, T2 and TIRM/FLAIR, all in low slice thickness (≤ 3 mm) to facilitate volumetric assessment and further include PWI and DWI to predict malignant transformation. Based on this survey, we cannot make any recommendation regarding the use of additional advanced imaging techniques (spectroscopy, fMRI, PET).

More than 75% adhered to follow-up imaging intervals of 12 to 24 weeks and 81% adjusted their imaging intervals depending on the initial growth rate of the tumor [21], which is in accordance with current guidelines [2]. It has been shown that the velocity of tumor expansion is a strong predictor of the patient’s prognosis [21]. However, no studies have addressed the question of how often MRI should be obtained during follow-up, especially with regard to the heterogeneity of DLGG. Although early postoperative FLAIR is known to overestimate the volume of residual tumor [22], most authors recommend an early postoperative MRI within 72 h [23‐25] to determine the extent of resection and visualize possible postoperative complications. Follow-up imaging intervals of 12–24 weeks are recommended with longer intervals for cases of “less aggressive” tumours [1, 13, 26]. The definition of “aggressiveness”, however, varies in the literature.

Only 15% of the centers reported to use RANO criteria for low-grade gliomas [13] thoroughly in all their DLGG cases for the interpretation of treatment response, while 46% of centers do so “most of the time”. Thus, 39% of centers do not routinely apply these criteria. RANO criteria include T1wo/w, T2/FLAIR, development of new lesions, clinical status, and steroid use in order to categorize treatment effects in complete response, partial response, stable disease, and progressive disease. Since DLGG constitutes a slowly progressive disease, these criteria were defined in order to achieve a standardized common ground for the definitions of treatment responses and endpoints in clinical trials. Measurement of tumor diameter has various shortcomings, not only the high intra- and interobserver variability [27], but also the known problem of head positioning during acquisition of the MRI [16, 17]. Then, assuming that many centers are performing volumetric assessment of tumors, how should treatment response be defined? Translating volume to diameter (D = (2 × V)^(1/3)) is a fundamental step. In contrast to the assessment of tumor diameters, the concept of volume-derived mean diameter overcomes the above mentioned confounding factors. Moreover, the curve showing evolution of mean diameter with time can be easily analyzed by applying a linear fit. Following the curve of the diameter as a function of time is a more sensitive way to monitor treatment response than applying RANO criteria [21, 28, 29]. Indeed, the major concern about RANO criteria is that pretreatment dynamics is not integrated in the definition of the different response categories. However, it seems obvious that putting down the tumor growth rate to 0 mm/year with chemotherapy, while its pretreatment value was 6 mm/year, should be interpreted as a response, whereas RANO criteria would interpret this as “stable disease”.

In contrast to this highly standardized protocol, our survey focused more on the clinical routine in an attempt to accurately reflect every day practice in the treatment of DLGG across Europe. Although the highly standardized approach provided by the RANO criteria is not systematically applied, our data demonstrate that at least the imaging studies, comply in most centers with the RANO-defined protocols.

Additionally, other factors should also be interpreted into our decision making, such as cognitive testing, psychological burden of disease, and seizure activity.

There is no consensus regarding the radiological malignant transformation in DLGG throughout Europe. Newly apparent contrast enhancement indicating breakdown of the blood brain barrier represents the classical sign of malignant transformation (MT) of these tumors [30], but preceding changes in advanced MRI investigations may allow identification of patients at risk up to 12 months earlier [31]. Simple measurement of growth rate [19, 32, 33] and integration of MR spetroscopy [18, 19, 34] are used routinely in most of the centers as predictors of tumor transformation. Both, proton- and phosphorus spectroscopy, available in numerous centers, have been proven to correlate with Ki67 and IDH1 mutation [35]. Perfusion measurements, and in particular the determination of relative cerebral blood volume (rCBV), seems to correlate with the vascularity determined at histopathological examination [31]. Arising of lactate resonance is predictive of the increase of rCBV up to 1.75, this parameter is predictive for a dramatic decrease of overall survival [36‐38], and are indicators of MT that may be identified months before apparent contrast enhancement. Likewise, ADC can be used for discrimination of tumor subtypes and raising suspicion of malignization [20]. Nevertheless, both techniques are only supported by a low level of evidence [20].

The use of amino acid PET (aaPET) was surprisingly variable between centers and countries. While Austria and Belgium performed PET in 100% of cases, none of the Dutch or French centers used PET at all. In contrast to patients with WHO grades III/IV gliomas, the evidence for aaPET to monitor patients with WHO grade II gliomas is limited [39, 40]. Most WHO grade II gliomas are nonenhancing with infiltrating tumor borders, and so several studies demonstrated the usefulness of aaPET in defining tumor extent or malignant transformation [41‐45]. This has been demonstrated and validated in series for 11C-MET, 18F-FET, and 18F-FDOPA PET [46]. Although MRI is the standard of care in following DLGG patients, its reliability in distinguishing tumor tissue from treatment effects is limited [47]. Transient blood–brain barrier alteration with contrast enhancement after radiotherapy with or without concomitant Temozolomide, for example, can mimic tumor progression. There are numerous reasons for the restricted application of PET: (1) While 18F-FDG is used routinely, access to aaPET is limited [46]. (2) Another obstacle to withhold patients and healthcare professionals from PET is limited reimbursement. In our survey, 9% of centers repeated aaPET scans during follow-up, even if they were initially negative. However, there are several groups advocating to perform aaPET in all cases of DLGG [48], notwithstanding the initial uptake behavior.

fMRI has repeatedly been shown to harbor a low specificity and sensitivity for any presurgical evaluation [49‐53]. Since the tumor itself impairs oxygenation levels in its surrounding fMRI is not a reliable surrogate marker for neuronal activity in DLGG patients [50, 51, 54]. Nevertheless, eloquence is thoroughly based on fMRI in as much as 16% of centers.

Yet, 80% of European centers prefer invasive intraoperative direct electrical cortical and subcortical stimulation over noninvasive procedures for the determination of eloquent cortex, which reflects commonly agreed practice and level of evidence [55‐57].

With the still small but increasing distribution of nTMS, some centers chose this technique for surgical decision making, which also illustrates the highly specialized nature of the enrolled centers. Although nTMS is a noninvasive modality, several reports proved not only the accuracy but also the feasibility of using it as a highly reliable tool for presurgical planning and intraoperative navigation [23, 58, 59] of primary motor functions. While non-invasive mapping can be reliably performed for primary motor functions (task-based fMRI, nTMS), this is currently not true for higher order functions (movement coordination, language, spatial consciousness, mentalizing, …) [60].

MEG was also rarely used. Nonetheless, despite its limited availability and high costs, its usefulness for presurgical planning and follow-up has repeatedly been reported [61‐63].

Irrespective of the used modality, noninvasive evaluation of eloquent function, especially if adjacent to or within the tumor is mandatory in order to identify the optimal time point for re-resection. This is even more relevant in the contrast of tumor-induced cortical reorganization potentially allowing gross total resection of previously unresectable tumors [64‐68].

In the near future, ongoing developments might be capable to achieve a change in imaging practice. Alongside with huge developments of machine learning using conventional MRI, which suppose the robustness of standardized sequences; an increased interest of metabolic multinuclear MR imaging and integration of multiparametric data into realistic metabolic-dynamic mathematical models is noticed.