Long-Term Survival of Patients With Melanoma With Active Brain Metastases Treated With Pembrolizumab on a Phase II Trial

Abstract

Purpose: Pembrolizumab is active in melanoma, but activity in patients with untreated brain metastasis is less established. We present long-term follow-up of pembrolizumab-treated patients with new or progressing brain metastases treated on a phase II clinical trial ( ClinicalTrials.gov identifier: NCT02085070).

Patients and methods: We enrolled 23 patients with melanoma with one or more asymptomatic, untreated 5- to 20-mm brain metastasis not requiring corticosteroids; 70% of patients had prior systemic therapy. Pembrolizumab was administered for up to 24 months. Brain metastasis response, the primary end point, was assessed by modified Response Evaluation Criteria in Solid Tumors (RECIST). Pretreatment tumors were analyzed for T-cell infiltrate and programmed death ligand 1.

Results: Six patients (26%) had a brain metastasis response. Eight patients (35%) did not reach a protocol evaluation scan and were unevaluable for brain metastasis response as a result of progression or need for radiation. Brain metastasis and systemic responses were concordant, with all ongoing at 24 months. The median progression-free and overall survival times were 2 and 17 months, respectively. Eleven patients (48%) were alive at 24 months. This included three unevaluable patients. One of these three patients had hemorrhaged, and two had symptoms from perilesional edema requiring radiosurgery, but all three patients remained on commercial pembrolizumab more than 24 months later. None of the 24-month survivors received subsequent BRAF inhibitors. Neurologic adverse events occurred in 65% of patients; all adverse events but one were grade 1 or 2. Three patients had seizures, which were treated with anticonvulsants. Most responders had higher pretreatment tumor CD8 cell density and programmed death ligand 1 expression, whereas all nonresponders did not.

Conclusion: Pembrolizumab is active in melanoma brain metastases with acceptable toxicity and durable responses. Multidisciplinary care is required to optimally manage patients with brain metastases, including consideration of radiation to large or symptomatic lesions, which were excluded in this trial. Two-year survival was similar to patients without brain metastasis treated with anti-programmed cell death 1 agents. Concordant brain and extracerebral responses support use of pembrolizumab to treat small, asymptomatic brain metastases.

FIG 1.

Characteristics of responses in the brain and extracerebral metastasis sites. (A) Best brain metastasis response by modified Response Evaluation Criteria in Solid Tumors (RECIST) and extracerebral response by RECIST version 1.1 in evaluable patients. Bars represent individual patients; blue bars and red bars represent changes in evaluable brain and extracerebral lesions, respectively. The amplitude of the bars demonstrates the percent change in tumor burden from baseline to the best response. We note that the patient on the left had progression in the brain and a mixed response in extracerebral sites with unequivocal progression of nontarget lesions. (B) Horizontal bars represent individual patients (evaluable or unevaluable) who remained on trial for ≥ 4 months. Bars are partitioned and color coded to indicate the time that patients were on study drug (pembrolizumab 10 mg/kg every 2 weeks), commercial pembrolizumab (2 mg/kg or 200 mg every 3 weeks), or off pembrolizumab. The timing of brain metastasis response, overall progression, death, or censoring is marked on the corresponding position of the bars. The dashed line represents the time the study ended (24 months).

FIG 2.

Kaplan-Meier curves for (A) progression-free survival (PFS) and (B) overall survival (OS). The dashed lines represent the time when the study ended at 24 months.

FIG 3.

Scatter plot of programmed death ligand 1 (PD-L1) levels versus CD8-positive area for individual patients. Values presented are the log transformation of 1 plus the value. Patients with biopsies from extracerebral sites are denoted by squares, and patients with brain biopsies are denoted by circles. Seventeen of the 20 specimens were collected before therapy. (A) Patients with progressive disease (PD) are denoted in blue, those with stable disease (SD) in red, and responders (either complete response [CR] or partial response [PR]) in orange, and three unevaluable (UE) patients are denoted in teal. Note that overall response is depicted in this figure given the small number of patients with evaluable specimens, including two patients who had response in extracerebral sites but were unevaluable for CNS response. (B) Patients alive at 24 months are denoted in orange, and patients dead at 24 months are denoted in blue.

FIG 4.

Quantitative immunofluorescent staining of programmed death ligand 1 (red) and CD8-positive T-cell infiltrating lymphocytes (green) in matched brain and extracranial sites in both (A and B) a responder and (C and D) a nonresponder, respectively. Nuclear compartment staining was defined by 4′,6-diamidino-2-phenylindole signal (blue).