Our results, based on serial blood samples collected prospectively, suggest that the evaluation of cfDNA concentration, in contrast to
TERTp mutation from plasmatic ctDNA, may give clinically relevant information in patients treated with RT-TMZ for glioblastoma. At baseline, we observed that the cfDNA level was associated with tumor neoangiogenesis based on MRI parameters (R-squared 0.32 regarding association between rCBV and cfDNA concentration). Our results also highlighted that cfDNA underwent significant variation during the treatment course, with a significant decrease after initial surgery or biopsy procedures and a significant increase in cases of PD during the TMZ maintenance phase. As yet, the role of cfDNA in patients treated for glioblastoma has not been clearly established. In a subgroup of 12 patients, Bagley et al. recently reported that cfDNA increased in cases of PD and remained stable in nonprogressive patients, similar to our findings [
7]. Moreover, the interest of cfDNA as a reflection of tumor burden was previously reported in a series showing that its level may correlate with radiological parameters. Taken together, these findings support the hypothesis that cfDNA is released from the vascularized part of the tumor [
20] and that this marker might be a useful tool for disease monitoring in the TMZ maintenance phase.
The evaluation of plasmatic ctDNA has been rarely investigated in prospective studies of patients with glioblastoma treated with RT-TMZ. In our work, only 2 cases with detectable circulating
TERTp mutation (n = 2/52, 3.8%) were observed using ddPCR, whereas 85% of the tumors were mutated, which was in accordance with previous studies [
11,
21]. The
TERTp mutation detection rate was similar to that of Juratli et al., showing a rate at baseline of 7.9% by nested PCR in 38 patients [
6]. ddPCR was used for
TERTp mutation detection, because this sequencing method is an ultrasensitive, fast and cost-effective tool, enabling its routine use in clinical practice. The lack of detected
TERTp mutations in plasma could be related to the short size of the ctDNA fragments (< 70 bp), which are not able to be detected by current sequencing methods [
22]. ddPCR is a sensitive method for detecting DNA fragments with amplicon sizes of 113 bp and 88 bp for
TERTp C228T and
TERTp C250T. Unfortunately, the use of the 88 bp amplicon size did not increase the sensitivity of ddPCR in our study. These results are to be balanced by the fact that some samples presented very few mutated droplets (samples considered as negative regarding our LOD) and that the quantity of DNA used for the 88 bp assay was less than that used for the 113 bp assay. Another point of discussion to increase detection rate would be to increase the supply of ctDNA. In our study, the patients had an average of 6 mL of whole blood per collection time, which appeared ethically acceptable. The question of increasing the volume of blood collected to increase the detection of ctDNA could be asked in a dedicated methodological study. Although the detection of circulating
TERTp mutation is a rare, our results also underlined that its detection may be of particular interest in two situations: extraneural metastatic dissemination, as previously reported in a case of metastatic ependymoma [
12], and the gliosarcoma subtype. Extraneural metastatic dissemination of gliomas is a rare, end-stage event, occurring in less than 2% of all cases [
23] and might be linked to blood–brain barrier rupture with hematogenous dissemination favored by first-line surgery [
24]. Unlike glioblastoma, gliosarcoma is characterized by its ability to invade the skull and disseminate systemically [
13]. It has been postulated that in cases of gliosarcoma, the blood–brain barrier has increased permeability, with a consequently greater release of ctDNA into the general blood circulation [
25]. In contrast to the ddPCR method, the use of targeted next-generation sequencing (NGS) would likely increase the ctDNA detection rate by identifying other recurrent altered genes [
7], but this did not correspond to our primary objective which was to identify a single biomarker of disease evolution. The clinical utility, as well as the sensitivity, of ctDNA detection in plasma by ddPCR and the NGS method require further characterization in larger cohorts, especially compared to CSF [
26,
27]. The choice of a non-selected population was made in order to explore the interest of the liquid biopsy concept in the daily-practice of patients with grade IV glioma (resection and biopsy, glioblastoma
IDHwt, glioblastoma
IDHmut, gliosarcoma). Regarding our results, a prospective study dedicated to the gliosarcoma or bulky
IDHwt glioblastoma subgroup would be of major interest.
Finally, the survival analyzes failed to show an impact of cfDNA, in contrast to the survival data reported by Bagley et al. [
7]. The survival results should be interpreted with caution due to the lack of power. Prospective and dedicated data are warranted to identify any prognostic significance of the cfDNA baseline concentration. An early cfDNA decrease after a diagnostic procedure should be investigated in a larger homogeneous cohort by separately analyzing resected tumor patients and bulky tumor patients. In addition, it would be interesting to explore the impact on survival of cfDNA kinetics before and right after RT-TMZ phase, before TMZ maintenance phase. The specific impact of corticosteroid therapy on the tumor release of cfDNA remains unclear despite a trend of association observed in our cohort at baseline (21 ng/mL vs. 14.4 ng/mL (
p = 0.1)). There is a high likelihood that the prednisolone equivalent dose and the tumor volume are confounding factors for cfDNA. During TMZ treatment, the cfDNA concentration continues to vary according to the tumor evolution—independent of the corticosteroid dose adjustment. Conversely, very high doses of corticosteroid delivered as a bolus could contribute to cfDNA release into the bloodstream and may explain the early cfDNA decrease between biopsy and the pre-RT-TMZ time point in this specific subgroup for whom no oncologic treatment was administered. As previously reported in other cancers, a transient spike of cfDNA may be observed right after tumor procedure [
28]. In our cohort we did not investigate this early cfDNA variation due to the lack of blood sample few days after diagnostic procedure. This early variation should be interesting to analyze regarding glioblastoma and biopsy procedure in further studies. No association between cfDNA and CRP, as inflammatory state marker, was observed. However, it would be interesting in further studies to analyze the relationship between cfDNA, corticosteroid, plasmatic inflammatory markers and white blood cell counts.