From 2013 to 2022, 73 patients were included in the final analysis out of the 133 patients identified in the database (Fig. 1)(Fig. 1B). The median age was 75 years (range 70–87), and the sex ratio was 1.9 (Table 1). A vast majority of the patients underwent biopsy-alone as the diagnostic procedure (n = 53/73, 72.6%). No significant differences were observed between the resected and unresected groups in terms of age (mean 74.8 versus 76.1, p = 0.215), CIRS-G score (mean score 4.7 versus 4.5, p = 0.78), or Karnofsky index score for those available (n = 31/70, 42.5%, mean 69% versus 77%, p = 0.138). Only 5 patients (6.8%) of the entire population had an oncogeriatric evaluation before treatment. Regarding treatment after the neurosurgical procedure, data were available for 74% of the population (n = 54/73). Among these patients, 79.6% (n = 43/54) had radiotherapy-based first-line treatment, and 46.3% (n = 25/54) received temozolomide, either concurrently with radiotherapy or as maintenance therapy. No patient had chemotherapy alone.

Description of the study. Schematic representation of the workflow of the study (A) and flow chart (B) of the population based on the database of the University Hospital of Rouen. The screened patients were aged 70 or over at the time of diagnosis (between January 1, 2013, and December 31, 2022)

In the entire population, 30 patients (41.1%) had at least one EGFR alteration, namely a mutation (n = 17, 23.3%), an amplification (n = 25, 34.2%) and/or a truncated gene (n = 10, 13.7%) (Table 1). No clinical differences at baseline or treatment schedule between the EGFRalt group and the EGFRwt group were observed (Table 1). The median overall survival in the entire population was 6.9 months [95% CI 5.3–8.4]. The overall survival rate at one year was 22% [95% CI 15–35%] and at two years was 4.2% [95% CI 1.4–13%]. The presence of EGFR alterations did not impact overall survival in the studied population: HR 0.97 [95% CI 0.6–1.57], p = 0.9. The median overall survival was 6.5 months [95% CI 5.3–9.3] in the EGFRalt group versus 7 months [95% CI 4.5–10] in the EGFRwt group, p = 0.9 (Fig. 2A). In detail, neither copy number variation (Fig. 2B), mutation (Fig. 2C), nor truncated gene (Fig. 2D) had an impact on OS. For the CNV analysis, HR was equal to 1.42 [95% CI 0.85–2.37], p = 0.2, for the copy neutral and 0.59 [95% CI 0.22–1.55], p = 0.3, for chromosome 7 polysomy compared to EGFRamp group. When pooling patients that carried EGFRamp glioblastoma (n = 25) and those with chromosome 7 polysomy (n = 5), a trend for a favorable outcome was observed compared to EGFR copy neutral in univariate analysis, but this was not confirmed when adjusting for other prognostic factors such as resection status (Table 2). Tumor resection was associated with a significant overall survival improvement: the median OS in the resected group (n = 20) was 11 months [95% CI 7.8–22] versus a median OS of 5.5 months [95% CI 4.6–7.8] in the unresected group (n = 53), log rank p < 0.0001. EGFR alteration status was not associated with survival outcome in the biopsy-alone group: HR 1.3 [0.72–2.32], p = 0.4. Eighty-eight percent of the EGFR mutations occurred in the extracellular domain of EGFR protein (Fig. 3) as previously described in the TCGA cohort [20]. The small number of mutations in our cohort did not allow to explore potential different prognostic impact of EGFR mutation’s location.

Kaplan–Meier curves according to the different EGFR alterations in the studied cohort (n = 73). The proportion of alive patients over time is presented according to the presence of at least one EGFR alteration in the tumor (A), the status of copy number variation (B), the presence of at least one mutation (C), and the presence of a truncated gene (D). Chromosome 7 polysomy is defined as a copy number gain of chromosome 7 higher than 2.5, based on the PI3KCG gene (7q22.3). Samples having both chromosome 7 polysomy and EGFR amplification were considered as EGFR-amplified glioblastoma. p-value is from the log-rank test

Schematic Representation of the EGFR Protein and the Location of the Altered Amino Acids Identified in the Studied Population (n = 73). Overall, 35 mutations were identified in 17 samples (23.3%) of the cohort. Of these mutations, 91.4% (n = 31/35) were missense mutations, one was a nonsense mutation, one was an insertion, and one was a splice variant. The mutations affecting the amino acids of the EGFR protein are represented. The pie plot by location represents the proportion of mutated samples within this amino acid among the entire cohort

Sixty-six patients were selected from the TCGA dataset, with a median age of 76 years (range 70–89) and a sex ratio of 1.2. Twenty-nine patients had glioblastoma with EGFR amplification (n = 29/66, 43.9%), while 12 patients had glioblastoma with EGFR mutation (n = 12/66, 18.2%). The median overall survival was 7.4 months [95% CI 4.9–11]. When combining the two cohorts, the one-year survival rate was 28% [95% CI 21–37%] and the two-year survival rate was 5.3% [95% CI 2.4–11.2%], Supplementary Table 1. Neither EGFR amplification nor EGFR mutation status influenced overall survival: the median survival was 8 months [95% CI 6.0–12] for glioblastoma with EGFR amplification compared to 5.9 months [95% CI 4.6–8.4] for EGFR copy-neutral glioblastoma, with a log-rank p-value of 0.22; and 7.8 months [95% CI 4.9–14] for glioblastoma with EGFR mutation versus 6.3 months [95% CI 4.9–8.4] for glioblastoma with wild-type EGFR, with a log-rank p-value of 0.18 (Fig. 4). These results were corroborated by univariate Cox regression analysis: hazard ratio (HR) of 1.26 [95% CI 0.87–1.82] for EGFR amplification and HR of 1.37 [95% CI 0.87–2.17] for EGFR mutation.

Kaplan–Meier curves stratified by EGFR amplification or mutation status in the studied cohort (n = 73) and the TCGA cohort (n = 66). Overall survival is displayed based on the type of alteration. The p-value is derived from the log-rank test

In the studied population, the in-patient rate was obtained for 60.3% (n = 44) of the entire cohort. Among this subgroup, almost half of the patients (n = 21/44, 47%) were hospitalized twice or more during active or palliative management. The three most frequent causes for hospitalization were: deterioration in general condition (35%), neurological impairment including deficit, intracranial hypertension, or seizures (26%), and treatment-related toxicity (19%). The in-patient rate was not associated with the EGFR alteration status of the tumor: 90% in the EGFRalt group (n = 18/20) versus 95.8% in the EGFRwt group (n = 23/24), p = 0.870.

When considering EGFR CNV as a continuous variable, the mean copy number gain was 37 copies (min. 2.82 – max. 111.55) in the EGFR gained glioblastoma population (EGFR amplification + chromosome 7 polysomy, n = 30). In the EGFRamp group (n = 25), all tumors harbored a copy gain higher than 10 copies. No correlation between EGFR CNV and clinical features was identified (Fig. 5).

Correlation plot between clinical, therapeutic features and EGFR copy number in the studied cohort (n = 73). EGFR copy number is set at 2 copies for non-amplified and non-polysomic samples. Tumor resection variable was transformed as binary variables. No correlation between those variables and EGFR copy number was identified. The only significant correlation was identified between overall survival (months) and tumor resection status (p < 0.01)

In the subgroup of patients with adjuvant treatment data available (n = 54), radiotherapy, regardless of the delivered dose, or temozolomide administration were identified as favorable prognostic factors: HR 0.55 [0.34–0.89], p = 0.015 and HR 0.15 [0.07–0.31], p < 0.001, respectively. The rate of patients receiving radiotherapy and/or temozolomide was equivalent between the EGFRalt subgroup and the EGFRwt subgroup: 70% versus 51.2% for radiotherapy, p = 0.171, and 48% versus 44.8% for temozolomide, p = 1.