Treatment patterns and clinical outcomes for patients with melanoma and central nervous system metastases: A real‐world study

Abstract Background Patients with melanoma and central nervous system (CNS) metastases have poor survival outcomes. We investigated real‐world treatment patterns and overall survival (OS) of patients with melanoma and CNS metastases. Methods A retrospective analysis utilizing a nationwide de‐identified electronic health record‐derived database was undertaken in patients diagnosed with advanced melanoma between January 2011 and September 2018. Patients with any visit ≤90 days of metastatic diagnosis and with confirmed CNS metastases were included. Results Of 3473 patients diagnosed with advanced melanoma, 791 patients with confirmed CNS metastases were identified and included in this analysis. Synchronous CNS metastasis (≤30 days of metastatic diagnosis) was associated with longer median OS than metachronous CNS metastasis (>30 days after metastatic diagnosis, 0.58 vs 0.42 years). Stereotactic radiosurgery (SRS) was the most common treatment (40.5%) alone or in combination with other local or systemic therapies, being more frequent in patients diagnosed in 2015+ versus 2011–2014 (44.1% vs 35.5%, respectively). The most common systemic treatment was immune checkpoint inhibitors (ICIs; 30.5%), predominantly anti‐cytotoxic T‐lymphocyte antigen 4 (CTLA‐4) alone (2011–2014) and anti‐programmed death‐1 alone or in combination with anti–CTLA‐4 (2015+). Median OS was longest in SRS‐treated patients (1.17 years) regardless of number of CNS metastases. Median OS for SRS‐treated patients increased from 0.83 years (2011–2014) to 1.75 years (2015+). In multivariable analysis, the effect of SRS remained significant after adjustment for sex, race, intracranial and extracranial disease burden, and timing of CNS metastases. Interaction testing to examine potential synergy between SRS/whole‐brain radiation therapy and ICIs found no significant interaction. Conclusions Despite advances in treatment, patients with melanoma and CNS metastases have poor survival outcomes. Prevalence of SRS increased over time and was associated with improved outcomes.


| INTRODUCTION
Among common solid tumors, melanoma has the highest risk of metastasis to the central nervous system (CNS). 1,2 Up to 75% of patients with metastatic melanoma will develop CNS metastases during their disease, 3 most within 3 years of melanoma diagnosis. 1 CNS involvement adversely affects quality of life and is historically associated with poor prognosis, with median overall survival (OS) of 4-6 months in patients with melanoma with CNS metastases. 1,3,4 Standard treatments for melanoma with CNS metastasis include whole-brain radiation therapy (WBRT) and stereotactic radiosurgery (SRS), alone or in combination, and cytotoxic chemotherapy. 5 Unfortunately, these treatments have not shown significant improvement in OS. New treatments approved from 2011 to 2015, including BRAF inhibitors (BRAFi) 6,7 and MEK inhibitors (MEKi) [8][9][10][11] for BRAF V600 mutation-positive tumors and immune checkpoint inhibitors (ICIs) [12][13][14] have improved treatment outcomes for some patients. Patients with unfavorable prognostic factors, including CNS metastases and elevated lactate dehydrogenase levels, often experience less benefit. [15][16][17] Furthermore, patients with CNS metastases have largely been excluded from registrational studies of novel treatments for melanoma. Thus, there remains substantial unmet need for treatment options that improve OS in this patient population.
Immune checkpoint inhibitors 18,19 and BRAF/MEK inhibitors [20][21][22][23] have demonstrated intracranial activity. This activity is of limited duration (BRAFi ± MEKi) 21 or is associated with a high rate of toxicities despite durable benefit (ICIs), 18,19,24 suggesting the need for alternative or combination treatments. Real-world data on treatment patterns and outcomes can inform treatment options and identify groups with high unmet needs, but data are lacking in the context of newer treatments. The objective of this study was to characterize real-world treatment patterns and survival of patients with melanoma and CNS metastases in the United States.

| Patients
This retrospective study utilized data from the Flatiron Health database (Flatiron, NY, NY), a nationwide longitudinal, de-identified database derived from electronic health record data. 25 During the study period, the de-identified data originated from ~280 cancer clinics (~800 sites of care). The patient cohort for this study comprised a random sample of eligible patients selected for enhanced manual chart review. De-identified patientlevel data included structured and unstructured data, curated via technology-enabled abstraction. 26 Institutional review board approval of the study protocol with a waiver of informed consent was obtained prior to study conduct.
Patients diagnosed with advanced melanoma from January 2011 through September 2018 (inclusive) were eligible for selection. Patients with melanoma were selected by International Classification of Diseases (Ninth and Tenth Revision) codes ICD-9 172.x or ICD-10 C43x or D03x. Patients with CNS metastases were selected and confirmed through review of clinical or pathologic records. Extracranial disease burden was captured by distinct anatomic sites and confirmed through review of medical records and pathologic reports. Synchronous versus metachronous CNS metastasis was defined using a cut-off of 30 days after first metastatic diagnosis to account for variability in local practice and access to care.
Patients with any visit with the Flatiron Health network ≤90 days of metastatic diagnosis (metastasis in any region) and confirmed to have CNS metastases at first metastatic diagnosis or at any time during follow-up were included. Of 3473 patients diagnosed with advanced melanoma, 791 patients who had a metastatic diagnosis date, a visit ≤90 days of metastatic diagnosis, and a CNS metastatic diagnosis were included in the current analyses ( Figure S1).

| Treatment
Initial treatment for CNS metastatic disease was defined as first therapy (local and systemic) started ≤90 days of CNS metastasis diagnosis date and any other therapies received ≤90 days of that first therapy. Treatments were classified as systemic (i.e., immunotherapy, targeted therapy, chemotherapy, other nonlocal therapies) or local (WBRT, SRS, other radiation therapy, craniotomy/metastasectomy). Systemic therapies were further categorized by therapeutic class. ICIs included anti-programmed death-1 (PD-1) or Conclusions: Despite advances in treatment, patients with melanoma and CNS metastases have poor survival outcomes. Prevalence of SRS increased over time and was associated with improved outcomes.

K E Y W O R D S
CNS metastases, melanoma, overall survival, radiotherapy, systemic treatment anti-programmed death ligand-1 (PD-L1) therapies (i.e., pembrolizumab, nivolumab, atezolizumab) and anticytotoxic T-lymphocyte antigen 4 (CTLA-4) (ipilimumab), alone or in combination. Targeted therapies included BRAFi (i.e., vemurafenib, dabrafenib, encorafenib) alone or in combination with MEKi (binimetinib, cobimetinib, trametinib). All patients were in mutually exclusive treatment groups based on a treatment hierarchy rule: evidence of any SRS; no SRS, any WBRT; no SRS or WBRT, any systemic therapy; and no SRS, WBRT, or systemic therapy, or evidence of any other therapy. Additional analyses were also performed to evaluate outcomes according to specific treatment groups defined by radiation therapy (SRS and/ or WBRT) and/or ICIs. The primary outcome was OS (i.e., time from diagnosis of CNS metastasis until death or last contact, estimated using the Kaplan-Meier method). Multivariable analyses were performed using Cox proportional hazards modelling to evaluate the impact of specified covariates on risk of death. The model included factors identified from descriptive comparisons (initial treatment within 90 days of CNS metastatic diagnosis, CNS metastatic burden, and synchronous vs metachronous CNS metastatic diagnosis) as well as established prognostic factors for which sufficient data were available from charts (sex, race, presence of liver metastases, and extracranial metastatic burden). Additionally, interaction testing was performed to examine potential synergy between SRS/WBRT and ICIs.  Table S1).

| Patients
Central nervous system metastases occurred early in most patients and survival outcomes remained stable, with shorter survival with higher intracranial tumor burden. Among the 791 patients with confirmed CNS metastases, synchronous CNS metastasis (≤30 days after first metastatic diagnosis) occurred in 522/791 patients (66.0%) and metachronous CNS metastasis (>30 days after first metastatic diagnosis) occurred in 269/791 patients (34.0%). Median time from CNS metastatic diagnosis to death or last contact was 4.9 months (IQR, 2.0-11.0), and was longer in patients with ≤3 versus >3 CNS metastases (7.
Immune checkpoint inhibitors were the most common systemic therapy used in combination with defined local therapies. Across treatment groups by treatment hierarchy rule, ICI use increased from 18
Stereotactic radiosurgery was associated with improved OS outcomes regardless of intracranial tumor burden; OS for patients with high intracranial tumor burden was shorter than for low intracranial tumor burden. Patients treated with SRS had the longest median OS for low (≤3 CNS metastases: 1.33 years [95% CI, 1.00-1.83]) and high intracranial tumor burden (>3 CNS metastases: 0.83 years [95% CI, 0.50-1.08]; Figure 1B,C). Among SRS-treated patients, 1-year OS rates were 57.1% (95% CI, 49.7-63.8) F I G U R E 1 Kaplan-Meier curves of overall survival by first treatment a received ≤90 days of diagnosis of CNS metastasis in (A) all patients with CNS metastases, (B) patients with ≤3 CNS metastases, and (C) patients with >3 CNS metastases. a Mutually exclusive groups based on treatment hierarchy rule. Abbreviations: CNS, central nervous system; met, metastasis; SRS, stereotactic radiosurgery; WBRT, whole-brain radiation therapy    Figure S3).

| DISCUSSION
Real-world data from US patients suggest that clinical outcomes for patients with melanoma and CNS metastases remain poor, particularly for patients with higher intracranial tumor burden. Initial SRS treatment had the most favorable survival outcomes and increased in prevalence over time. Median OS for patients treated with SRS improved in 2015+ versus 2011-2014, which may reflect improvements in technology, administration, and availability. Patients who are candidates for SRS have lower number of metastases and smaller tumor burden, and are likely to have better OS. There was increased use of combination ICIs, and previous reports have suggested potential synergistic interaction between SRS and immunotherapy or targeted therapy, which may improve OS. 20,[27][28][29] Use of anti-PD-(L)1 alone and in combination with anti-CTLA-4 increased from 1.7% to 21.9% and 0.0% to 17.9%, and use of BRAFi + MEKi combination therapy increased from 0.0% to 4.5% from 2011-2014 to 2015+ among SRS-treated patients, respectively. While use of WBRT and BRAFi monotherapy decreased slightly from 2011-2014 to 2015+, it remained higher than expected, suggesting a degree of sub-optimal treatment in the real-world setting.
Patients with no evidence of treatment had the least favorable survival outcomes; median OS (3 months) was similar to untreated patients with melanoma and CNS metastases in the SEER-Medicare database (2.1 months), indicating that patients who were not candidates for a therapeutic intervention were likely close to end of life. 30 32 Further comparisons between our study and the Canadian study are limited due to treatment category and time period differences. An additional report from Australia from January 2011 through December 2014 demonstrated that patients who received SRS combined with systemic therapy had improved outcomes compared with other therapies, including SRS alone. 33 In our study, patients who received initial treatment with SRS combined with ICIs had slightly better OS than those who received SRS alone. Both SRS and ICIs were independently associated with improved OS outcomes in multivariable analysis. However, interaction testing to examine potential synergy between SRS/WBRT and ICIs did not demonstrate a significant effect.
In the current analysis, synchronous CNS metastasis diagnosis was associated with improved survival outcomes for patients with <3 CNS metastases. This may support routine CNS imaging for early detection and treatment of CNS metastases, especially in those with low intracranial burden. In the Canadian study, use of brain magnetic resonance imaging for surveillance or metastatic restaging increased over time and was associated with better survival outcomes in both patient-level and regional-level analyses, suggestive of a survival benefit with earlier CNS metastases detection. 32 Our analysis provides valuable insight into real-world clinical practice with longitudinal evaluation reflecting recent advances in melanoma therapy. However, we acknowledge several limitations. Most patients were treated at community rather than academic practices and do not reflect potential differences in patient populations and treatment patterns between these settings. Electronic health record databases, while less costly and time consuming compared to primary data collection, may contain missing or incomplete data. Patients may have received noncaptured treatment outside of the Flatiron Health network. Additionally, the absence of links to claims data prevented reliable capture of information on use of oral medications such as corticosteroids that may influence survival outcomes. Based on validated algorithms developed for the prediction of benefit from SRS and selection of eligible patients, intracranial tumor burden was categorized as ≤3 versus >3 CNS metastases. Abstraction methodology limitations prevented a more granular assessment of the impact of CNS tumor burden, particularly for patients with >3 CNS metastases. Analyses were limited to treatments received ≤90 days after diagnosis and did not consider additional treatments received following this time.

| CONCLUSION
Despite expansion of effective therapies for patients with metastatic melanoma, patients with CNS metastases, particularly with greater tumor burden, continue to have poor survival outcomes, representing an area of high unmet need. Real-world data from the current study highlight the increased prevalence of SRS over time and associated improved outcomes, potentially as it is used more consistently along with ICIs. The benefit of concurrent SRS and ICIs remains uncertain and requires confirmation in wellcontrolled clinical studies.