It is now well recognized that DLGG might exhibit spatial heterogeneities [15]. New imaging techniques enable to detect such heterogeneities. In particular, dynamic ¹⁸F-FET-PET allows to identify foci of decreasing time-activity curves, that correlate with histological grade [16]. Similarly, the role of perfusion imaging to detect such hot spots is still investigated. Therefore, although this has not be proven, it can be anticipated that removal of such malignant foci within an otherwise DLGG could improve oncological benefit of surgery, even though the entire glioma was not completely resected.

Finally, epigenomic and genomic landscapes and their evolution in the course of glioma growth are being progressively deciphered [17–20]. Such studies evidenced a branching pattern rather than clonal evolution. Importantly, the different branches are located in different spatial spots in the tumor [19, 20]. Thus, it can be anticipated that a better modeling of this genomic evolution could allow to identify a hot spot within the tumor bearing a branching clone of higher aggressiveness. Surgical removal of this specific area could result in enhanced oncological benefit, even if the resection is partial.

It has been shown that awake surgery with intraoperative neurological and neuropsychological testing and monitoring greatly reduces the functional risk of DLGG surgery [21]. Nevertheless, it should not be forgotten that immediate postoperative cognitive evaluations report a deterioration in most of patients, with a (quite) complete recovery after 3 months of intensive rehabilitation. It is acknowledged that this full cognitive recovery is made possible by the plasticity of distributed networks [8]. But the sine qua non condition for network normalization by postoperative plasticity is the surgical preservation of a minimal core of long-range connectivity [22–24], including but not limited to superior, middle and inferior longitudinal systems, inferior fronto-occipital fasciculus, cingulate fasciculus, frontal aslant tract, and callosal fibers. Such axonal pathways are essential to maintain the small-world topology of brain networks [25]. Therefore, these functional white matter tracts should be imperatively identified by direct subcortical electrostimulation mapping in all DLGG patients, whatever the tumor location, and they should serve as limits of surgical resection [26].

Considering that the neuronal areas per se of this core of connectivity are safely controlled thanks to intraoperative monitoring, small deep arterial infarcts remains the major source of damage of the long-range connectivity. Statistical map of diffusion hypersignal with decreased ADC (an MRI signature of acute ischemia) after glioma resection is lacking, as well as the functional correlates of such small strokes depending on their topography. It can be hypothesized that any stroke encompassing the minimal common brain [23] (i.e. the basic core of associative connectivity), would impact high level cognitive functions.

Moreover, for a similar insult of the connectome, the degree of recovery will vary from one patient to another, depending on their level of plasticity reserve. Hence, estimating the functional risk of a given patient means evaluating its plasticity reserve, which is currently not straightforward. First of all, a slight deterioration in some of the preoperative cognitive scores reflects that neuroplasticity has already spent much of its reserve. Interestingly, preoperative infiltration by glial cells of the connectivity of the ventral stream has been shown to correlate with a decrease of semantic fluency scores [27]. In other words, preoperative infiltration by the glioma of the minimal core of connectivity can hamper the reshaping processes of networks reconfiguration, precluding a fully effective functioning. This preoperative infiltration of the connectome has also to be taken into account regarding postoperative plasticity. Let us consider for example the effect of ILF resection on lexical access [28]: the degree of recovery could differ depending on the degree of preoperative infiltration of the other associative tracts involved in this function (inferior fronto-occipital fasciculus, long direct segment and short posterior segment of arcuate fasciculus [22]). This would be especially true in a fast growing glioma, because the axonal dysfunction induced by glioma infiltration has an impact on the process of neuronal synaptic reweighting underlying network normalization. In the same vein, postoperative radiation therapy on a residue left in the connectome could seriously compromise the chances of recovery [29].

Finally, age is a major factor of plasticity potential: as age increases, plasticity decreases. Surprisingly, there are no study in the literature reporting the influence of age on cognitive recovery after awake surgery of DLGG. In our experience, we observed that after 55 years of age, recovery took longer (between 8 and 18 months) in comparison to younger patients (for example, median time to work resumption of 4 months in [30]).

Since postoperative cognitive evaluations are performed on a routine basis, our knowledge of the impact of surgery on high-level functions has greatly improved, leading to a better preoperative personalized estimation of functional risk [31]. To this end, the importance of the preoperative patient interview cannot be overemphasized. Patient should clearly formulate his wishes regarding which kind of functions should be primarily preserved. This will depend on its way of life, including profession, hobbies, and social interactions [32]. For example, it has been recently recognized that proper names retrieval is frequently impaired after left dominant temporal lobectomy [33, 34]. In any profession requiring a high level of proper names retrieval (for example when you need to deal with a lot of client names), such deficit precludes work resumption—while for some other patients, such a deficit would not affect quality of life. Another example is provided by fine motor functions : it has been reported that resection of SMA glioma without identifying networks of motor control [35–37] results in long lasting trouble in bimanual coordination [30, 38]. Again, such deficit goes unnoticed for some patients, while it would dramatically impact the professional life of a musician or a tennis player. In a similar way, the ecological consequences of the resection of optic radiations (causing an hemianopia) is not of the same importance whether the patient absolutely needs to drive or not [7]. In the same state of mind, the level of cognitive control (i.e. executive functions) can be set preoperatively with the patient, and intraoperative task can be personalized accordingly. Several tasks requires a high cognitive load, including but not limited to, dual task, working memory tasks, or TMT part B. Preserving such high-level cognitive capabilities is probably more important for a manager than for a cleaning lady. The same principles apply to emotional functions, like mentalizing. The read the mind in the eye task allows to identify the low-level network of mentalizing [39], and it is anticipated that preserving this network is of utmost importance for patients with a profession requiring a high level of empathy, like medical doctor.

Finally, there is a controversy regarding the possibility to offer a complete resection that would come together with a high risk of neurological deficit [40–42]. Resection of tumors involving the anterior perforated substance (APS), usually insular or paralimbic gliomas [43], is a paradigmatic example of this ethical issue. Complete resection of the APS comes with a very high risk of hemiplegia. Surgeons agreeing with the moto “primum non nocere” would definitely not go beyond the functional responses elicited in the external capsule covering the APS, thus giving up to resect the tumor part within the APS. In this tumor location, some other surgeons plaid for informing patients about the two options (safe resection with suboptimal oncological benefit versus maximizing oncological benefit at the price of hemiplegia) and offer them to get involved in this decision, that is, to accept (or not) severe permanent deficit in order to increase the extent of resection. However, it is questionable whether the patient can objectively imagine how he could enjoy life with an hemiplegia and/or a marked cognitive deficit—because cutting the temporal stem to reach the APS will also induce definitive higher-order disturbances (in addition to the motor deficit related to the injury of the lenticulo-striate arteries when removing the APS itself), especially due to a damage of the inferior fronto-occipital fasciculus, even in the right hemisphere [44].

Functional improvement after resection of DLGG have been reported [30, 45, 46]. Among other explanations, this could be due to the relief of mass effect or to the better reorganization of brain functional networks after tumor removal [47]. Moreover, the positive impact of resection on epilepsy status should not be overlooked [48]. It has been suggested that resection of temporo-mesial structures could be offered for epilepsy control even if not infiltrated by the glioma [49]. This functional benefit (which might be of utmost importance given that seizure control is the key parameter allowing patients to keep their driving license) has to be weighted with the functional risk of verbal memory decline, which is a well known consequence of temporo-mesial structures removal.

Importantly, the functional risks of the surgery should also be weighted in comparison to the spontaneous evolution of the DLGG. Indeed, as reported above, in absence of any surgery, there is a progressive cognitive decline all along the natural course of the disease, related to the infiltration of the minimal common brain. Of note, in a old series, it was claimed that with an endpoint at 4 years, surgery was cognitively more risky than wait and watch management [50]. However, this result might be out of date, considering that surgery was not performed under awake mapping at that time. In addition, even assuming that this would be still valid nowadays (which is very unlikely taken into account the favorable neurological and neuropsychological outcomes when using intraoperative awake mapping [45, 46, 51–54]), a 4-years time-point is not enough to draw any conclusion, considering that median survivals have now reached more than 10 years. In fact, any surgery that would delay the infiltration of the connectome should help to delay cognitive decline. Hence, (supra-)complete resection while preserving the minimal core of connectivity provides a functional advantage compared to wait and see. On the contrary, if a residue is left within the infiltrated connectome, the functional advantage of surgery is only provided by the oncological advantage, that is by delaying malignant transformation. Indeed, it has been shown that the residue will grow at the same speed as before surgery [55], meaning that in this situation, the connectome infiltration will not be delayed by the surgery compared to wait and see.

Current data on the real oncological benefit of radiation therapy are scarce. Indeed, it has been shown that in terms of survival, there is no benefit of early versus late radiation therapy in DLGG [63].

From a functional point of view, long-term adverse effects have been well described [64]. This study found a significant cognitive decline 7–10 years after irradiation, compared to a well matched population of DLGG patients without any irradiation. Currently, the factors that would allow to predict these late adverse effects in order to improve our evaluation of onco-functional benefit of radiation therapy are poorly known: are they related to age and cardiovascular parameters? to the volume of irradiation? to the location of the irradiated area, especially within or outside the minimial common brain—i.e. not only irradiation of the hippocampi but also irradiation of the functional white matter tracts? Interestingly, a recent series showed that diffusion tensor imaging can predict cognitive function deficit following partial brain radiotherapy for DLGG, in particular with an increased radial diffusion at the end of radiotherapy able to significantly predict decline in verbal fluency 18 months after irradiation [29].

Last but not least, short-term adverse effects are poorly described in the context of DLGG. Only a very recent paper reported a 20% rate of pseudoprogression [65], with a median delay of 12 months (range 3–78 months). In this study, no clinical symptoms were attributed to these pseudoprogressions, but one can fear that such reactions would be harmful whenever the irradiation targeted a residual tumor left by the surgeon for functional reasons.

These combined data about lack of survival benefit of early treatment and late onset of adverse effects grounded the recommendation to postpone radiation in DLGG whenever survival is expected to be greater than 7–10 years [66].

Beyond the onco-functional balance intrinsically related to a given treatment, the choice of a treatment modality should be reevaluated in light of its interaction with associated treatments in the whole sequence. In this part, we propose to review some of these interactions.