A total of 146 patients treated with single fraction salvage SRS for HGG were included in the study. A summary of patient demographics is shown in Table 1. The median age of patients was 58 years (range 20 to 83 years) and 60 years (range 21–84) at time of diagnosis and time of SRS, respectively. There were 83 males (56.9%) and 63 females (43.2%). The majority of patients were diagnosed with histologic WHO Grade IV gliomas (92.5%, 135/146 patients). The IDH-mutation status was wild-type in 94 patients (64.4%), IDH-mutant in 15 patients (10.3%), and unknown in 37 patients (25.3%). Of the patients included, 71.9% identified as non-hispanic white, 17.2% as hispanic, 7.5% as black, and 3.4% as asian. The median Karnofsky Performance Status (KPS) score at time of SRS treatment was 60 (range 20–90). With regards to the resective surgery, a gross total resection was achieved in 58 patients (39.7%), a near total resection in 17 patients (11.6%), subtotal resection in 64 patients (43.8%), biopsy without resection in 5 patients (3.4%), and thermal ablation in 2 patients (1.4%). The median time from diagnosis to salvage SRS treatment was 12.2 months (range 2.99 to 175.8 months). Of the patients for whom data on the use of bevacizumab was collected, 95 (65.1%) received bevacizumab during their overall treatment course, 12 (8.2%) did not receive bevacizumab, and 39 (26.7%) are unknown.

In our cohort of 146 patients, 410 targets were treated with SRS. A single focus of recurrence was treated in 53 patients (36.3), 2 foci in 27 patients (18.5%), 3 in 28 patients (19.2%), 4 in 17 patients (11.6%) and greater than 4 lesions in 21 patients (14.4%). The median number of lesions treated at each session was 2 (range 1–11 lesions). The median number of times patients underwent treatment was 1 (range 1–5). Of all treated lesions, 10.7% had IDH mutations (44 out of 410 targets in 15 patients).

The median enhancing tumor targeted for each focus was 3.3 cm3 (range 0.0–85.5). The median target volume treated was less than 2.5 cm3 in 190 lesions (46.34%), between 2.5 cm3 and 7.0 cm3 in 86 lesions (20.98%), and greater than 7.0 cm3 in 134 lesions (32.68%). The median total dose was 18 Gy (range 10–24 Gy) to the 50% isodose line. The SRS dose was less than 18 Gy for 174 treated lesions (42.44%), between 18 Gy and 20 Gy for 135 lesions (32.93%), and greater than 20 Gy for 101 lesions (24.63%). A summary of the characteristics of treated lesions is shown in Table 2.

The median follow-up after salvage SRS was 8.2 months. A total of 61 (41.8%) patients continued with progression following first SRS treatment, with a median PFS of 8.28 months (95% confidence interval [CI]: 6.9–10.3). PFS at 6 and 12 months were 65.8% (95% CI: 57.4–73.0) and 37.3% (95% CI: 29.3–45.3), respectively (Fig. 1). Median OS from time of diagnosis was 26.6 months (95% CI: 23.8–34.4) and 12.4 months (95% CI: 10.2–15.4) following first SRS (Fig. 2). Median OS for patients with and without progression after first SRS treatment was 16.6 months (95% CI: 13.0-22.1) and 9.33 months (95% CI: 6.8–12.2) respectively. One patient was lost to follow-up and removed from further analysis.

Progression-free survival following first SRS. (SRS = stereotactic radiosurgery, CI = confidence interval)

Overall survival following first SRS. (SRS = stereotactic radiosurgery, CI = confidence interval)

Of the total number of lesions treated (N = 410), 142 treated lesions had imaging demonstrating failure. There were 113 local failures (79.5%), 2 distant failures (1.4%), and 27 concomitant local and distant failures (19%). Of the local-only failures, 60 (53.1%) were considered in-field, 51 (45.1%) were marginal, and 2 (1.8%) were regional. A summary of the patterns of failure is shown in Table 3, and exemplar images are shown in Figs. 3 and 4. Median time to local failure after first SRS treatment was 17.1 months (95% CI: 13.3–24.1). Local failure-free survival at 1 and 2 years were 65.1% (95% CI 55.1–73.4) and 40.4% (95% CI: 28.8–51.7) respectively (Fig. 5).

Example MRI images demonstrating different failure patterns (in-field, marginal, regional, and distant failure) as well as local control. (SRS = stereotactic radiosurgery)

Example pre-SRS treatment MRI images from a patient in whom multiple synchronous recurrent foci were treated with SRS (left) with corresponding T2 FLAIR sequences (right)

Local failure-free survival after first SRS. (CI = confidence interval)

Table 4. shows the clinical and demographic variables predicting local failure following salvage SRS treatment. Age ≥ 60 years at SRS (HR = 2.29, 95% CI: 1.18–4.48) was significantly associated with local failure, with a median of 4.8 months to local failure in patients ≥ 60 years vs. 6.14 months in patients < 60 years. KPS ≥ 70 at time of first SRS (HR = 0.44, 95% CI: 0.24–0.80) and non-white race (HR = 0.60, 95% CI: 0.38–0.97) was also significantly associated with local failure (median time to local failure of 11.3 months in patients with KPS ≥ 70 vs. 3.6 months in patients KPS < 70, and 10.6 months in non-white race vs. 4.6 months in those who identified as white). SRS dose ≥ 20 Gy (HR = 0.34, 95% CI: 0.16–0.76) was an additional predictor of local SRS failure with longer median time to failure (9.7 months vs. 3.7 months in those who received 18–19 Gy and 6.4 months in those who received < 18 Gy). After adjusting for age, race, and KPS at recurrence, SRS dose ≥ 20 Gy remained significantly associated with longer time to failure (HR = 0.36, 95% CI: 0.17–0.75). The cumulative incidence of local failure was highest amongst patients ≥ 60 years treated with an SRS dose < 20 Gy (Fig. 6). The use of bevacizumab was not associated with local failure (P > 0.05).

Incidence of local failure divided by age and SRS dose. The cumulative incidence of local failure was highest amongst patients ≥ 60 years treated with an SRS dose < 20 Gy

Table 5. shows the clinical and demographic variables associated with salvage SRS failure location. Non-white race (HR = 5.55, 95% CI: 1.55–19.94) and SRS dose ≥ 20 Gy (HR = 5.62, 95% CI = 1.27–24.92) were associated with a more distant pattern of failure. After adjusting for KPS at diagnosis and race, SRS dose ≥ 20 Gy was no longer associated with failure location (HR = 3.70, 95% CI = 0.82–16.64). Sex, glioma grade, and use of bevacizumab were not associated (P > 0.05) with SRS failure or pattern of failure. No grade 4 or 5 toxicity was observed.

Sub-group analysis was performed on patients in whom IDH wild-type mutation status was confirmed (i.e. excluding patients with IDH mutation or unknown). A total of 93 patients with 248 targets were included. The median target enhancing volume was 3.15 cm3 and median total dose was 18 Gy to the 50% isodose line. Forty-eight (51.6%) patients developed progression following SRS. PFS at 6 and 12 months were 64.1% and 40% respectively. Median OS from time of diagnosis was 22.2 months and 12.94 months following first SRS. Analysis of failure pattern demonstrated 68 (77%) local failures (53% in-field, 44% marginal, 2% regional), 2 (2%) distant failures, and 18 (21%) concomitant local and distant failures. Local failure-free survival at 1 and 2 years was 42.2% and 15.6%, respectively.

Analysis of clinical and demographic variables predicting local failure following salvage SRS treatment found age < 60 years (HR = 0.37) and KPS at time of first SRS ≥ 70 (HR = 0.38) to be associated with longer time to local failure. Median time to local failure was 7.4 months in patients ≥ 60 years vs. 9.6 months in patients < 60 years and 17.4 months in patients with KPS ≥ 70 vs. 5.8 months in patients with KPS < 70. SRS dose, use of bevacizumab, and race were not statistically significant. Additionally, there were no clinical and demographic variables found to be statistically significantly associated with salvage SRS failure location.