Background Immune checkpoint inhibitors (ICI) have anti-cancer activity in selected patients with central nervous system (CNS) metastases. However, the benefit of ICIs in patients with leptomeningeal metastases (LMM) is unknown. We hypothesized that pembrolizumab would lead to CNS responses in patient with LMM from solid tumors, and that genomic and immunologic features of the cerebrospinal fluid (CSF) may identify biomarkers of LMM response.

Methods We undertook a single-center investigator-initiated phase 2 trial of pembrolizumab in patients with LMM from solid tumors. Eligible patients had radiologic (>3mm on MRI) or cytologic (+CSF cells) LMM and ECOG PS 0-1. Pembrolizumab was administered IV at 200mg q3W until disease progression or unacceptable toxicity. The primary endpoint was CNS response (complete response=CR, partial response=PR or stable disease=SD) after 4 cycles, defined radiologically/cytologically/clinically. Radiologic response was assessed by RECIST v1.1 and irRC. Secondary endpoints were CNS-progression-free survival (PFS), overall survival (OS) and safety. Baseline and serial CSF samples were assessed by tumor-derived DNA aneuploidy assay (t-DNA), 16-color flow cytometry and multiplex cytokine analysis.

Results Thirteen of a planned 18 patients were treated between 04/2017-12/2019. The study closed early due to poor accrual. Median age was 57 years (range 22-79); 54% were female. The majority of patients had tumors not traditionally responsive to ICI (62%: hormone-receptor+ breast carcinoma=39%; high-grade glioma=23%), while 38% had ICI-responsive tumors (NSCLC=23%, head&neck carcinoma=8%, cutaneous squamous carcinoma=8%). CNS response was observed in 38% of patients (95% CI 13.9-68.4%). Two patients achieved durable CRs (cutaneous squamous carcinoma=1, OS 3+yrs; MET-exon14+ NSCLC=1, OS 9 mos.), 1 PR (7.7%, OS 6 mos), and 2 SDs (15.4%) in the CNS. Median CNS-PFS and OS were 2.9 mos (95% CI: 1.3-NR) and 4.9 mos (95% CI: 3.7-NR), respectively. There were no unacceptable safety signals. Sensitivity for LMM detection by t-DNA was 84.6% (95% CI: 54.6-98.1%), and 46.2% (95%CI: 19.2-74.9%) by cytopathology. Pre and on-therapy CSF cytokine analysis showed complete responders clustered together, while progressors clustered differently.

Conclusions Patients with LMM from solid tumors have a dismal prognosis and limited treatment options. In this phase 2 trial, we identified an impressive 38% CNS response rate for pembrolizumab in patients with LMM, deep and durable responses in selected patients with ICI-responsive tumors, and that pembrolizumab was well-tolerated. CSF t-DNA may be more sensitive for detection of LMM than cytopathology, and immunologic subsets of ICI-response based on cytokine profiles warrant further study. These data support investigation of pembrolizumab in larger populations with LMM.

Trial Registration NCT03091478

Ethics Approval The study was approved by John’s Hopkins University’s Institututional Ethics Board, approval number J1655

Consent All participants provided informed consent as per the study protocol