Clinical outcomes of melanoma brain metastases treated with stereotactic radiosurgery and anti-PD-1 therapy, anti-CTLA-4 therapy, BRAF/MEK inhibitors, BRAF inhibitor, or conventional chemotherapy

systemic treatment regimens and SRS technique

Nivolumab or pembrolizumab comprised the anti-PD-1 therapy regimen. Seven patients (33%) were treated with pembrolizumab and 14 patients (67%) were treated with nivolumab. Patients treated with nivolumab received 3 mg/kg q 2 weeks or 2 mg/kg pembrolizumab q 3 weeks. The anti-CTLA-4 therapy, ipilimumab, was administered over 4 infusions every 3 weeks dosed to 3 mg/kg. All patients receiving the BRAF/MEKi regimen or BRAFi alone harbored the V600E mutation. The BRAF/MEKi regimen consisted of daily doses of dabrafenib 150 mg twice a day (b.i.d.) and trametinib 2 mg daily administration. Patients receiving BRAFi alone received vemurafenib 960 mg b.i.d. In patients receiving concurrent BRAF/MEKi or BRAFi alone with SRS, patients were instructed to hold the drug 2–3 days before and after SRS. The majority of patients treated with conventional chemotherapy received carboplatin and paclitaxel (n = 8; 40%). The remaining conventional chemotherapy cohort consisted of three patients treated with dacarbazine (15%), three patients temozolomide (15%), two patients carboplatin and paclitaxel followed by temozolomide (10%), two patients cisplatin, dacarbazine, and vinblastine (10%), one patient carboplatin and dacarbazine (5%), and one patient cisplatin, vinblastine, and temozolomide (5%).

Treatments were delivered with the BrainLab Novalis Classic linear accelerator (LINAC) with 6 MV photons. These methods have been previously described [11]. All MBMs were treated with SRS in a single session.

follow-up

Patients in this study were followed with clinical examinations and MRI imaging at 2–3 month intervals. There were no differences in clinical or radiographic follow-up intervals between treatment groups. Images were assessed alongside two neuro-radiologists at our institution (JAA and DGS). The primary end point of this study was distant MBM control. Distant MBM control was defined as freedom from development of new MBM or leptomeningeal disease outside of the irradiated field. Local MBM failure was defined as at least a 20% radiographic increase in the size of the previously irradiated area in accordance with RECIST criteria [12]. RECIST criteria have been used previously by our group [10, 11, 13] and was determined to be more established than RANO criteria [14] at the time of analysis. Secondary end points in this study included local MBM control, systemic PFS (sPFS), as well as OS from the date of SRS and the date of intracranial diagnosis and neurotoxicity.

statistical analysis

Statistical analyses were carried out using JMP 11 (SAS Institute Inc., Cary, NC). The local and distant MBM control rates, as well as OS, were calculated from the date of MBM diagnosis or radiation treatment to the date of progression or death using the Kaplan–Meier (KM) method, with the log-rank test used to test differences. To test differences between cohorts, the Kruskal–Wallis, Pearson's χ², and Fisher's exact tests were used when appropriate. The Cox proportional hazard model analyses were carried out using univariate analysis (UVA) and multivariate analysis (MVA). Variables that showed significant effects on UVA (P < 0.1) were included in the MVA.

patient and lesion characteristics

A total of 96 patients treated to 314 MBMs over 119 SRS sessions were identified. Patient and lesion characteristics for the group are summarized in supplementary Table S1, available at Annals of Oncology online. The median follow-up of patients after MBMs diagnosis was 7.4 months (range: 0.1–46.2 months).

The majority of SRS sessions were treated concurrently or after systemic treatment (n = 72; 61%) and the majority of lesions were treated with 24 Gy (n = 128; 41%). The median gross tumor volume (GTV) was 0.14 cm³ (range: 0.008–9.48 cm³).

patient and lesion characteristics among groups

Patient and lesion characteristics among the five systemic treatment groups are displayed in Table 1. Overall, the five groups were well balanced in terms of single or multiple metastases treated per patient (P = 0.46), gender (P = 0.85), age (P = 0.27), diagnosis-specific graded prognostic assessment (P = 0.51), use of steroids at the time of SRS (P = 0.25), number of systemic organs affected by metastases (P = 0.11), time from diagnosis of distant metastatic disease to time of SRS (P = 0.40), and GTV (P = 0.26). Differences were noted in the median SRS dose used per lesion among groups (P = 0.005). Thirty-eight percent and 20% of patients were BRAF mutated in the anti-PD-1 and anti-CTLA-4 cohorts, respectively, while all patients treated with BRAFi had mutant tumor (P = 0.0001). Owing to the era of diagnosis, BRAF mutation status was largely unknown for patients treated with conventional chemotherapy. In addition, there was a difference between groups in administration of previous systemic treatments (P = 0.001).

distant and local MBM control

KM distant MBM control rates at 6 and 12 months across all treatment sessions were 36% and 18%. Significant differences were noted among the groups in distant MBM control rates following SRS, P = 0.008 (Figure 1). Six- and 12-month distant MBM control rates were 61%/38% (anti-PD-1 therapy), 26%/21% (anti-CTLA-4 therapy), 53%/20% (BRAF/MEKi), 30%/8% (BRAFi), and 15%/5% (conventional chemotherapy) following SRS. Distant MBM control UVA and MVA is detailed in Table 2. Treatment with BRAF/MEKi or anti-PD-1 therapy was found to be significant for distant MBM control on UVA. This remained significant on MVA (chemotherapy/anti-PD-1 HR 3.1 [95% confidence interval (CI) 1.5–6.6, P = 0.001]) and chemotherapy//BRAF/MEKi HR 2.1 (95% CI 1.1–4.4, P = 0.03)]. No significant differences were noted in distant MBM control with management of BRAFi (P = 0.27) or anti-CTLA-4 therapy (P = 0.26) over conventional chemotherapy on MVA. Combining immunotherapy groups (anti-PD-1 and anti-CTLA-4), this cohort continued to have improved distant MBM control compared with chemotherapy [chemotherapy/immunotherapy HR 2.1 (95% CI 1.1–3.6, P = 0.02)] on MVA, while targeted therapy (BRAFi and BRAF/MEKi) did not [chemotherapy/targeted therapy HR 1.2 (95% CI 0.71–2, P = 0.51)].

KM local MBM control rates at 6 and 12 months were 89% and 83%, respectively. There were no significant differences among the groups in local MBM control rates, P = 0.25. Dose (P = 0.63), prior systemic treatment (P = 0.57), and number of lesions treated per patient (P = 0.42) were not predictive of local MBM failure, supplementary Table S2, available at Annals of Oncology online.

systemic progression-free survival

The median sPFS from the date of SRS was 3.4 months (range: 0.47–45.9 months). Actuarial sPFS rates are displayed in supple-mentary Figure S1, available at Annals of Oncology online. Six- and 12-month sPFS rates following SRS were 41%/41% (anti-PD-1 therapy), 36%/27% (anti-CTLA-4 therapy), 58%/39% (BRAF/MEKi), 29%/12% (BRAFi), and 20%/5% (chemotherapy) (P = 0.04). UVA and MVA analysis of sPFS from the date of SRS is displayed in supplementary Table S3, available at Annals of Oncology online. Treatment with anti-PD-1, anti-CTLA-4, or BRAF/MEKi significantly improved OS on both UVA and MVA when compared with chemotherapy, MVA 2.7 (95% CI 1.4–5.7; P = 0.005), 2.5 (95% CI 1.3–5.0; P = 0.007), and 2.5 (95% CI 1.2–5.5; P = 0.02), respectively.

OS from SRS and cranial metastases diagnosis

The median OS for all patients in the study was 8.9 months (range: 0.47–48 months) from the date of SRS and 10.5 months (0.5–48.2 months) from the date of cranial metastases diagnosis. Actuarial OS rates by treatment group from the date of SRS are displayed in Figure 2. Six- and 12-month OS rates following SRS were 76%/48% (anti-PD-1 therapy), 68%/41% (anti-CTLA-4 therapy), 83%/65% (BRAF/MEKi), 52%/24% (BRAFi), and 60%/10% (chemotherapy) (P = 0.01). Similar differences were noted in OS from the date of cranial metastases diagnosis with 6- and 12-month OS rates of 81%/66% (anti-PD-1 therapy), 84%/50% (anti-CTLA-4), 83%/75% (BRAF/MEKi), 71%/29% (BRAFi), 70%/15% (chemotherapy) (P = 0.004). In those patients that died (n = 73; 76%), imaging at last follow-up before death was assessed. The percentage of patients demonstrating progression in the brain alone, systemic alone, or both brain and systemic progression in the anti-PD-1 therapy, anti-CTLA-4 therapy, BRAF/MEKi, BRAFi, and chemotherapy cohorts was 16%/33%/50%, 19%/19%/63%, 70%/0%/30%, 47%/20%/33%, 25%/15%/60%, P = 0.10, respectively.

UVA and MVA analysis for OS from SRS is displayed in sup-plementary Table S4, available at Annals of Oncology online. Treatment with anti-PD-1, anti-CTLA-4, or BRAF/MEKi significantly improved OS on both UVA and MVA when compared with chemotherapy, MVA 3.4 (95% CI 1.6–7.2; P = 0.0009), 3.1 (95% CI 1.5–6.4; P = 0.002), and 2.4 (95% CI 1.1–5.3; P = 0.02), respectively.

neurotoxicity

Symptomatic radionecrosis requiring surgical evacuation 12 months following SRS was noted in one mesial frontal lesion (GTV 0.93 cm³) treated with 18 Gy SRS 1 month before initiation of ipilimumab. Additionally, one right cerebellum lesion treated ∼2 weeks before ipilimumab initiation was thought to be radiographic radionecrosis 17 months following 21 Gy SRS and required steroids for symptomatic improvement of headaches. One right frontal lesion (GTV 0.49 cm³) treated with concurrent vemurafenib underwent surgical resection revealing radionecrosis 18.5 months following 21 Gy SRS. One right frontal lesion (GTV 3.7 cm³) treated with SRS concurrent with pembrolizumab revealed post-treatment hemorrhage and radiation necrosis on surgical pathology 2.5 months post-SRS.

One patient each in the anti-PD-1, anti-CTLA-4, and BRAFi groups experienced grade 2 headaches following SRS each managed with steroids. Two patients treated with conventional chemotherapy experienced grade two headaches following SRS requiring steroid management.