Survival and treatment outcomes in patients with leptomeningeal disease from metastatic melanoma

Abstract

Background

Melanoma leptomeningeal disease (LMD) has a poor prognosis. However, the management of patients with advanced melanoma has evolved with time, including those with LMD. We reviewed a large cohort of melanoma LMD patients to assess factors associated with survival.

Methods

Retrospective clinical data was collected on patients diagnosed with LMD at MD Anderson Cancer Center from 2015 to 2020. Overall survival (OS) was determined from LMD diagnosis to date of death or last follow-up. The Kaplan–Meier method and log-rank test were used to estimate OS and to assess univariate group differences, respectively. Multivariable associations of survival with variables of interest were determined using Cox proportional hazards regression models.

Results

A total of 172 patients were identified. The median age at LMD diagnosis was 53 (range 20–79) years, and all patients had radiographic evidence of LMD on magnetic resonance imaging of either brain or spine. In total 143 patients previously received systemic therapy (83%), with a median of 2 prior treatments (range 0–5). 81 patients (47%) had concurrent uncontrolled systemic disease and 80 patients (53%) had elevated serum LDH at the time of diagnosis. With a median follow-up of 4.0 months (range 0.1–65.3 months), median OS for all patients from LMD diagnosis was 4.9 months. Patients (n = 45) who received intrathecal therapy or systemic immunotherapy for LMD had a median OS of 8.0 months and 10.2 months, respectively. On multivariable analysis, decreased performance status, positive CSF cytology, elevated LDH, and whole brain radiation were associated with worse OS.

Conclusions

Despite many advances in therapeutic options, the outcomes of melanoma patients with LMD remains poor. However, a subset of patients appears to derive benefit from LMD-directed treatment.

One of the most common complications of advanced melanoma is the development of central nervous system (CNS) metastasis.¹ Leptomeningeal disease (LMD) is a subset of CNS disease that affects up to ~15% of patients with metastatic melanoma.² The reported incidence of melanoma parenchymal CNS metastases is rising likely due to improved imaging techniques, more diligent surveillance, and longer survival of patients with advanced disease.^2–4 Despite notable advances in the treatment of metastatic melanoma and parenchymal brain metastases, LMD survival remains poor.^1,3–7 Our prior retrospective analysis of 178 melanoma patients diagnosed with LMD between 1999 and 2015 revealed a median overall survival of 3.5 months with no significant improvement over that time period.⁷ Due to the poor prognosis and significant neurologic morbidity associated with LMD, patients with this diagnosis have traditionally been excluded from the vast majority of clinical trials. As such, the development of novel clinical trials to treat melanoma LMD has been limited, despite the unmet need to improve survival for this patient population. Results from recent prospective clinical trials suggest that systemic therapy, intrathecal therapy and/or their combination may prolong survival in melanoma patients with LMD.^8–11 Since the approval of the ipilimumab and nivolumab combination therapy by the US Food and Drug Administration (FDA) in 2015 for patients with metastatic melanoma, it has become the first systemic therapy to achieve durable response in melanoma patients with brain parenchymal metastases.^12–14 Our recent analysis of patients with parenchymal brain metastasis showed a significant improvement in overall survival (OS) since 2015.¹⁵ Thus, we conducted a retrospective chart review analysis of melanoma patients diagnosed with LMD from 2015 to 2020 at our institution and identified contemporary prognostic factors in this specific patient population. This work will add to our understanding of this heterogeneous patient population and might identify which patients will benefit most from further LMD-directed treatment.

Methods

Under a research protocol approved by the institutional review board of MD Anderson Cancer Center (MDACC), the database of melanoma informatics, tissue resource, and pathology core (Melcore; institutional melanoma clinical research database) was searched to identify and collect information from records of melanoma patients who were diagnosed with LMD between 6/11/2015 and 12/31/2020. None of the patients diagnosed in 2015 were included in the previously reported analysis.⁷

The cohort included patients with a diagnosis of LMD based on cytopathological analysis of cerebrospinal fluid (CSF) and/or magnetic resonance imaging (MRI) of the brain and/or spine.^16–19 Of note, no information on the number of CSF assessment by total lumbar puncture was collected, and no information on nodular versus linear LMD was included due to the large degree of variability in physicians recognizing or defining nodular LMD.²⁰ Clinical characteristics collected for all patients included age, sex, primary tumor characteristics, date of LMD diagnosis, therapies received (before and after LMD diagnosis), Eastern Cooperative Oncology Group (ECOG) performance status (PS), non-CNS disease status, and serum lactic acid dehydrogenase (LDH). The status of non-CNS disease was based on review of contrast computed tomography (CT) scan of the chest/abdomen/pelvis (CAP) or a body positron emission tomography (PET) scan within 30 days of LMD diagnosis. Uncontrolled non-CNS disease at LMD diagnosis was defined as progressive or new sites of metastatic disease when compared with previous imaging; controlled disease was defined as stable disease showing no evidence of radiographic progression or new sites of metastatic disease. OS time was computed from the date of LMD diagnosis to the date of last known vital status. Patients who were alive at the end of the study period were censored as of the date of the last follow-up. The Kaplan–Meier method was used to estimate OS and differences between groups were assessed using the log-rank test. The association between OS and groups of interest was determined using Cox proportional hazards regression models. Measures that occurred after LMD diagnosis (eg, LMD-directed treatment) were included in the Cox models as time-dependent covariates. All statistical analyses were performed using SAS 9.4 for Windows. All statistical tests used a significance level of 5%. No adjustments for multiple testing were performed.

Results

Patient Characteristics

Between 2015 and 2020, 172 melanoma patients with LMD were seen at MD Anderson Cancer Center and were included in the analysis. Baseline characteristics of the patients are shown in Table 1. The median age at LMD diagnosis was 53 years (range: 20–79 years). The majority of patients were male (n = 103, 60%). The ECOG performance status was 0 in 47 patients (27%). Most patients had a history of primary cutaneous, non-acral melanoma (54%; n = 93), and 113 patients (66%) had a BRAF mutation. Serum LDH level above the institutional normal limit was observed in 80 patients (53%) at the time of LMD diagnosis.

Table 1.

Characteristics of Patients With Melanoma LMD

Age at initial melanoma diagnosis (median, in years)	49 (range: 17–78)
Age at LMD diagnosis (median, in years)	53 (range: 20–79)
		n	%
Sex	Male	103	60
	Female	69	40
Melanoma subtype	Cutaneous	93	54
	Unknown primary	34	20
	Acral lentiginous	14	8
	Mucosal	9	5
	Primary CNS	7	4
	Uveal	5	3
	Unclassified	5	3
	No report at MDACC	5	3
Mutation status	BRAF	112	65
	NRAS	22	13
	BRAF and NRAS	1	1
	GNAQ	6	3
	KIT	1	1
	Other mutation	12	7
	Unknown mutation status	3	2
	No mutation	15	9
CSF cytology	Positive	44	26
	Negative	33	19
	Atypical cells	6	3
	No analysis performed	89	52
MRI brain radiographic findings	LMD positive	128	74
	Inconclusive for LMD on initial assessment	23	13
	LMD negative	20	12
	No MRI brain	1	1
MRI spine radiographic findings	LMD positive	69	40
	LMD negative	56	33
	No MRI spine	47	27
State of parenchymal brain metastases	Uncontrolled	119	70
	Controlled No disease present	29 24	17 14
State of systemic (non-CNS) disease	Uncontrolled	81	47
	Controlled—no disease present	55 36	32 21
Neurologic symptoms	No	46	27
	Yes	117	68
	Unknown	9	5
ECOG status at LMD diagnosis	0	47	27
	1	43	25
	2	28	16
	3	13	8
	4	7	4
	Missing	34	20
LDH elevated above institutional upper	No	70	41
normal limit at LMD diagnosis	Yes	80	47
	Missing	22	13
Any previous systemic treatment	No	29	17
	Yes	143	83
	Immunotherapy	123	72
	Targeted therapy	90	52
	Chemotherapy	27	16
	Bio chemotherapy	11	6
	Interferon	9	5
Any previous CNS radiation	No	67	39
	Yes	105	61
Time from prior CNS radiation to LMD diagnosis (median, in months)	3 (range: 0–73)
Previous surgical resection of brain metastases	No Yes	106 66	62 38
LMD directed treatments	Any treatment	144	84
	Radiation	92	53
	Whole brain RT alone	14	10
	Spinal XRT alone	5	3
	SRS alone	3	2
	Intrathecal	45	26
	Chemotherapy	30	17
	IMT	58	34
	Targeted therapy	60	35
	IMT + TT	30	17
	Chemo + IMT	8	5
	Chemo + TT	6	3
	Chemo + IMT + TT	1	1
Median number of LMD directed treatments	2 (range: 0–5)

Abbreviations: Chemo, chemotherapy; CNS, central nervous system; CSF, cerebrospinal fluid; ECOG, Eastern cooperative oncology group; IMT, immunotherapy; LDH, lactic acid dehydrogenase; LMD, leptomeningeal disease; MRI, magnetic resonance imaging; SRS, stereotactic radiosurgery; TT, targeted therapy.

Over 2/3rd of the patients (n = 127) had a history of parenchymal brain metastases prior to LMD diagnosis. Most of these patients had received CNS radiation (n = 105; 61%), which included whole brain radiation (WBRT) in 17 patients (10%), stereotactic radiosurgery (SRS) in 74 patients (43%), and spinal radiation in 16 patients (9%). The median time from the last CNS radiation to LMD diagnosis was 3 months (range: 0.03–73 months). Sixty-six patients (38%) had prior craniotomy, but only 4 patients (6%) had resection in the posterior fossa.²¹ Prior to LMD diagnosis, 143 patients (83%) received systemic therapy for metastatic melanoma, which included targeted therapy in 90 patients (52%), checkpoint inhibitor-based immunotherapy in 123 patients (72%), and chemo- or biochemotherapy in 38 patients (22%).

The median time from primary melanoma diagnosis to LMD diagnosis was 32.3 months (range: 0.03–332.4 months). The median time interval between diagnosis of metastatic melanoma and diagnosis of LMD was 13.3 months (range: 0.03–110.4 months) and the median time from parenchymal brain metastasis to LMD diagnosis was 4.2 months (range: 0.0–-78.6 months).

MRI of the brain showed evidence of LMD in 128 patients (74%) and MRI of the spine was positive for LMD in 69 patients (40%). Forty-five patients (26%) had radiographic evidence of LMD on both brain and spine MRI. At the time of LMD diagnosis, 136 (79%) patients had concurrent non-CNS disease, which was uncontrolled in 81 patients (60%). Concurrent parenchymal brain metastases were observed in 148 patients (86%), which were uncontrolled in 119 patients (80%). Thirty patients presented with de novo parenchymal brain metastases and concurrent LMD as their first CNS involvement. Twenty-four patients (14%) had LMD only.

Of the 83 patients (48%) who had CSF cytology analysis at the time of LMD diagnosis, 44 patients (53%) had positive CSF cytology, and 6 patients (7%) had “atypical” cells which could not be confirmed as melanoma. Among the CSF positive patients, 35 had neurological symptoms (data for 2 patients was missing), and 39 had a positive MRI brain for LMD. In the 4 patients without evidence of LMD in the brain, the spine MRI confirmed LMD. Data on opening pressure during lumbar puncture was available in only 55 of the patients with CSF cytology analysis and was elevated in only 13 patients (ranging from 20 to 55 cm H₂O). At the time of LMD diagnosis, 117 patients (68%) had neurological symptoms. A detailed summary of the most reported symptoms (headache, cranial nerve palsy, cognitive impairment, and others) is provided in Figure 1.

Figure 1.

Most common symptoms reported in patients with neurological symptoms. Data might be not complete due to the retrospective nature of this analysis.

With a median follow-up of 4.0 months (range: 0.1–65.3 months), 31 patients (18%) were still alive at the end of time frame of the study. The median OS from LMD diagnosis was 4.9 months (95% CI, 3.4–6.5).

The majority of patients in the cohort (n = 144; 84%) received at least one LMD-directed treatment following LMD diagnosis, (median = 2; range: 0–5 treatments). Of these treated patients, 92 (64%) received radiation therapy: WBRT only (n = 14, 10%), spinal radiation only (n = 5, 3%), stereotactic radiosurgery only (n = 3, 2%), or radiation therapy with other systemic therapies (n = 70, 49%). The median time from LMD diagnosis to first CNS radiation therapy was 0.3 months (range: 0.03–34.3 months). Following LMD diagnosis, initial therapy with systemic therapy was used in 122 patients (71%) and consisted of: checkpoint inhibitor-based immunotherapy in 86 patients (70%); targeted therapy with BRAF, BRAF/MEK, or MEK inhibitors in 74 patients (61%); and chemotherapy in 38 patients (3%), with temozolomide used in 28 patients (23%) and other chemotherapy used in 10 patients (8%). Forty-five patients (26%) received intrathecal (IT) agents; the most frequently used were nivolumab in 29 patients (17%) (in conjunction with clinical trial NCT03025256), followed by interleukin-2 (IL-2) in 23 patients (13%). The median OS from LMD diagnosis was 7.8 months for all patients who received systemic therapy; 10.2 months for those receiving checkpoint-inhibitor-based immunotherapy, 8.0 months for patients receiving targeted therapy, and 6.5 months for those receiving chemotherapy. The median OS for patients (n = 45) who received IT agents was 8.0 months.

Patient and Treatment Characteristics Associated With OS From LMD Diagnosis

Table 2 presents the univariate analysis of associations of patient characteristics and OS from LMD diagnosis. The median survival time for all patients in the cohort was 8 months (Figure 2A). Patients with performance status of ECOG 0 had longer OS (median, 8.0 months) compared to patients with ECOG 1–4 (median, 3.1 months; hazard ratio 1.64; 95% CI,1.10–2.45; P = .015). The absence of non-CNS disease or presence of controlled non-CNS disease was associated with significantly longer survival (median, 6.8 months) when compared with uncontrolled non-CNS disease (median, 3.4 months; HR 1.50; 95% CI,1.07–2.09; P = .018) (Figure 2B). Serum LDH above institutional normal limit was associated with significantly shorter OS (median, 3.2 months) when compared with normal serum LDH levels (median, 8.0 months; HR 2.24; 95% CI,1.54–3.26; P < .001). The absence of parenchymal brain metastases or the presence of controlled parenchymal disease was associated with longer OS (median, 7.8 months) when compared with the presence of uncontrolled parenchymal brain metastases (median, 4.4 months; HR 1.52 [1.03–2.22]; P = .033). Furthermore, positive CSF cytology at LMD diagnosis was associated with shorter OS (median, 6.4 months; HR 2.08 [1.18–3.66]; P = .011) (Figure 2C). Lastly, patients with LMD-only disease had better OS (median, 11.0 months) than those with LMD concurrent with non-CNS disease (median, 4.4 months; HR 0.47; 95% CI,0.24–0.90; P = .023) (Figure 2D).

Table 2.

Overall Survival and Patient Characteristics—Univariate Analysis

Measures	n	Median OS (95% CI) in months	P-value	Hazard Ratio (95% CI)	P-value
Sex
Male	103	5.5 (3.4, 7.8)	0.94	Ref
Female	69	3.9 (2.8, 6.4)		1.01 (0.72, 1.42)	.94
BRAF mutation
Mutant	113	5.1 (3.4, 7.8)	0.77	Ref
No BRAF	41	3.5 (2.4, 7.5)		1.14 (0.77, 1.70)	.51
Unknown mutation status	3	18.6 (2.8, 18.6)		0.62 (0.15, 2.51)	.50
No mutation	15	5.6 (1.4, 14.5)		0.91 (0.50, 1.65)	.75
Performance status ECOG
0	47	8.0 (5.2, 13.8)	<.001	Ref
1	43	5.5 (3.1, 13.0)		1.06 (0.65, 1.73)	.80
2	28	2.7 (2.0, 3.9)		2.25 (1.35, 3.76)	.002
3	13	1.0 (0.6, 1.9)		3.25 (1.65, 6.43)	<.001
4	7	1.6 (0.3, 3.1)		4.52 (1.98, 10.32)	<.001
Missing	34	5.5 (3.2, 7.9)		1.78 (1.09, 2.91)	.022
Performance status ECOG
0	47	8.0 (5.2, 13.8)	.028	Ref
1–4	91	3.1 (2.5, 4.1)		1.64 (1.10, 2.45)	.015
Missing	34	5.5 (3.2, 7.9)		1.73 (1.06, 2.82)	.029
Primary site
Cutaneous	93	4.5 (3.2, 7.8)	.15	Ref
Unknown primary	34	5.5 (3.9, 8.0)		0.88 (0.57, 1.37)	.58
Acral lentiginous	14	2.8 (1.4, 5.1)		1.23 (0.67, 2.27)	.50
Mucosal	9	1.9 (0.9, 6.5)		1.63 (0.78, 3.38)	.19
Primary CNS	7	18.6 (2.0, NE)		0.34 (0.12, 0.93)	.035
Uveal	5	2.6 (0.7, 10.9)		1.64 (0.60, 4.51)	.34
Unclassified	5	3.5 (2.3, 18.2)		0.96 (0.39, 2.38)	.93
No report at MDACC	5	3.4 (1.1, 7.9)		1.83 (0.73, 4.55)	.19
CSF cytology
Negative	33	13.1 (8.0, 15.4)	<.001	Ref
Positive	44	6.4 (3.4, 8.0)		2.08 (1.18, 3.66)	.011
Atypical cells	6	5.1 (1.5, 25.2)		2.05 (0.77, 5.47)	.15
No analysis	89	3.1 (2.5, 3.8)		3.18 (1.92, 5.26)	<.001
Radiographic evidence of LMD
Brain only	83	5.6 (3.3, 7.8)	.34	Ref
Spine only	24	3.9 (2.0, 13.8)		0.86 (0.52, 1.45)	.58
Brain and spine	45	4.4 (2.8, 6.5)		1.26 (0.85, 1.86)	.26
Neurological symptoms
No	46	6.5 (3.2, 13.1)	.68	Ref
Yes	117	4.1 (3.2, 6.3)		1.19 (0.80, 1.76)	.39
Missing	9	5.5 (1.3, 14.6)		1.19 (0.57, 2.48)	.65
State of parenchymal brain metastases
Controlled/absent	52	7.8 (2.8, 10.3)	.031	Ref
Uncontrolled	119	4.4 (3.4, 5.6)		1.52 (1.03, 2.22)	.033
Systemic non-CNS disease
Controlled/absent	91	6.8 (3.9, 9.9)	.017	Ref
Uncontrolled	81	3.4 (2.6, 5.1)		1.50 (1.07, 2.09)	.018
Serum LDH > ULN
No	70	8.0 (5.2, 13.0)	<.001	Ref
Yes	80	3.2 (2.3, 4.1)		2.29 (1.57, 3.34)	<.001
Missing	22	3.5 (2.1, 6.7)		2.22 (1.27, 3.86)	.005
LMD only
No	159	4.4 (3.3, 5.6)	.020	Ref
Yes	13	11.0 (2.8, 24.5)		0.47 (0.24, 0.90)	.023
Any previous systemic therapy
No	29	6.4 (4.9, 11.0)	.17	Ref
Yes	143	4.1 (3.2, 5.6)		1.37 (0.87, 2.15)	.17
Any previous CNS radiation
No	69	5.1 (3.4, 6.7)	.71	Ref
Yes	103	4.4 (3.2, 8.0)		0.94 (0.67, 1.31)	.71
Previous surgical resection
No	106	4.1 (3.1, 5.7)	.18	Ref
Yes	66	5.6 (3.3, 8.0)		0.79 (0.56, 1.11)	.18
LMD directed treatments (Yes vs. No)a
Any treatment				0.64 (0.41, 1.01)	.053
> 2 treatments				0.56 (0.38, 0.81)	.002
Radiation				1.26 (0.91, 1.77)	.17
Whole brain RT alone				2.92 (1.64, 5.19)	<.001
Spinal XRT alone				1.65 (0.67, 4.04)	.28
Intrathecal				0.79 (0.52, 1.21)	.29
Chemotherapy alone				1.33 (0.85, 2.07)	.21
IMT alone				0.84 (0.58, 1.22)	.37
Targeted therapy alone				1.03 (0.72, 1.49)	.86
Chemo + IMT				0.98 (0.45, 2.12)	.96
Chemo + TT				1.45 (0.62, 3.36)	.39
IMT + TT				0.94 (0.60, 1.49)	.80

^aIncluded in the model as a time-dependent covariate.

Abbreviations: CI, confidence interval; Chemo, chemotherapy; CNS, central nervous system; CSF, cerebrospinal fluid; ECOG, Eastern cooperative oncology group; IMT, immunotherapy; LDH, lactic acid dehydrogenase; LMD, leptomeningeal disease; NE, not estimated/not reached; ref = reference group; RT, radiation therapy; TT, targeted therapy; ULN, upper limit normal.

Figure 2.

(A) Kaplan–Meier estimates of overall survival for all LMD patients. The median survival time was 8 months. (B) Kaplan–Meier estimates of overall survival for LMD patients according to systemic disease status. Patients with controlled or absent (n = 91) systemic disease survived for 6.8 months. Patients with uncontrolled systemic disease (n = 81) had a median survival time of 3.2 months, (P = .011). (C) Kaplan–Meier estimates of overall survival of LMD patients according to CSF cytology. Patients with positive CSF cytology (n = 44) survived for 6.4 months. Patients with negative CSF cytology (n = 33) had a median survival time of 13.1 months, (P = .008). (D). Kaplan–Meier estimates of overall survival of LMD patients according to LMD status. Patients with LMD only disease (n = 13) survived for 11.0 months. Patients with concurrent other systemic disease with LMD (n = 159) had a median survival time of 4.1 months, (P = .019).

No single LMD treatment modality was significantly associated with improved OS from LMD diagnosis. However, patients who received more than 2 treatment modalities had significantly better survival (HR 0.56; 95% CI,0.38–0.81; P = .002). Further, treatment with WBRT was significantly associated with worse OS from LMD diagnosis (HR 2.94 [1.64–5.19]; P < .001).

Multivariable analysis was performed (Table 3) with all factors with P < .1 on univariate analysis. Patients with ECOG 1–4 had worse OS compared with patients with ECOG 0 (HR 1.63; 95% CI,1.06–2.51; P = .026). Positive CSF cytology also remained significantly associated with worse OS (HR 2.58; 95% CI,1.41–4.71; P = .002). Serum LDH above the institutional normal limit was associated with worse OS (HR 1.92; 95% CI,1.27–2.90; P = .002). Finally, WBRT remained significantly associated with worse survival (HR 2.29; 95% CI1.24–4.24; P = .008).

Table 3.

Overall Survival and Patient Characteristics—Multivariable Analysis

Measures	Hazard Ratio (95% CI)	P-value
Performance status ECOG
0	Ref
1–4	1.63 (1.06, 2.51)	.026
Missing	0.95 (0.52, 1.75)	.87
CSF cytology
Negative	Ref
Positive	2.58 (1.41, 4.71)	.002
Atypical cells	1.96 (0.69, 5.52)	.20
No analysis	2.79 (1.62, 4.81)	<.001
State of parenchymal brain metastases
Uncontrolled	Ref
Controlled/absent	0.94 (0.59, 1.48)	.78
Systemic non-CNS disease
Uncontrolled	Ref
Controlled/absent	1.02 (0.68, 1.54)	.92
Serum LDH > ULN
No	Ref
Yes	1.92 (1.27, 2.90)	.002
Missing	2.13 (1.05, 4.33)	.036
LMD only
No	Ref
Yes	0.66 (0.30, 1.45)	.30
>2 LMD treatment (Yes vs. No)a	0.48 (0.33, 0.72)	<.001
Whole brain RT alone (Yes vs. No)a	2.29 (1.24, 4.24)	.008

^aIncluded in the model as a time-dependent covariate.

Abbreviations: CI, confidence interval; LAD, lactate dehydrogenase; LMD, leptomeningeal disease;.

ref = reference group; RT, radiation therapy.

Discussion

LMD is one of the most challenging and devastating complications of advanced cancer, and historically, the survival of these patients has been extremely poor.²² The development and approval of multiple immune and targeted therapies in recent years have resulted in dramatic improvement and outcomes of patients with metastatic melanoma, including in patients with brain parenchymal metastasis.^12–14,23 Our recent comprehensive analysis of patients with melanoma brain metastasis showed significant improvement in the outcome and factors associated with survival in patients with brain parenchymal metastasis.¹⁵ To address the outcomes and factors associated with survival in patients with melanoma LMD in the current era with available systemic immune and targeted therapies, in this study, we aimed to assess the impact of current treatment options for LMD on survival; this study represents the largest single center cohort reported to date. The main finding of our study is that although there has been significant improvement in clinical outcomes for patients with metastatic melanoma, including in patients with parenchymal brain metastasis since the FDA approval of the combination therapy ipilimumab and nivolumab in 2015, this progress does not extend to patients with melanoma LMD. Our study showed that there has been comparatively meager improvement in OS for melanoma patients with LMD since 2015. Combined with our previous analysis of a large cohort of melanoma patients diagnosed with LMD from 1999 to 2015, these results illustrate the continued need for new and more effective therapeutic approaches for this patient population, particularly patients who present with factors that are associated with shorter OS.

The diagnosis of LMD is typically based on CSF cytopathology and/or contrast-enhanced MRI of the neuroaxis.^2,4,24 To date, the gold standard for diagnosis of LMD remains CSF analysis, but this modality has significant limitations due to the overall low sensitivity (50–60%) of the first sample of CSF.^16,22 Due to this limitation, recent studies have evaluated the use of advanced CSF methods including circulating tumor DNA and proteomic analysis of CSF, to aid in diagnosis and evaluate treatment responses. However, these methods are not currently routinely performed or available.^25–28 Multivariable analysis showed that positive CSF cytology was associated with a worse OS compared with patients with negative CSF cytology. This may allude to higher tumor burden in CSF positive patients driving this difference in outcome. Notably, no CSF analysis was also significantly associated with worse outcome. Possible explanations for this observation are that some patients declined too fast and/or at the time of diagnosis, the MRI of the brain and/or spine was sufficient for the presence of LMD, and these patients did not undergo lumbar puncture for CSF analysis. In line with our prior report, ECOG ≥ 1 and elevated LDH (which is a surrogate for systemic burden) remained significantly associated with poor OS.⁷

As summarized in Table 4, LMD survival remains poor, and often is still measured in weeks to a few months.^7,29,30 While most of these studies included only a relatively small number of patients, a few demographic features were similar across them. In general, the patient population was relatively young, with the median age ranging from 52 to 58. Male gender was predominant in all reports, and BRAF mutation was noted to be higher than in the average metastatic melanoma setting in 3 out of the 5 studies shown. The median OS is slightly more favorable in the current study and may be explained by all patients being treated at a high-volume, tertiary cancer center with access to contemporary melanoma therapies, as well as intrathecal therapies. Furthermore, multivariable analysis showed that patients undergoing multiple treatments (>2) had better OS. Notably, treatment with WBRT was associated with worse survival. It is important to note that it does not necessarily indicate that WBRT directly worsens outcome, rather since the role of WBRT is palliative relief and patients dispositioned for WBRT likely had higher disease burden and neurological symptoms requiring urgent intervention, these patients may have worse outcomes.⁷ In comparing our current results to our previous retrospective study, even though median survival is higher in the current study, approximately 5 months compared to 3.5 months in previous (2009–2015 cohort) study, it is still clear that the survival has not significantly improved in the last decade. In terms of specific factors associated with LMD outcome, both studies aligned on the impact of functional status, which is expected. Additionally, both studies noted that systemic disease status (radiographic or elevated LDH) at time of LMD diagnosis significantly impacted survival. A notable difference between the current and previous analysis is the impact of previous surgical resection on LMD survival. Both clinical cohorts (2009–2015) and current (2015–2020) included similar number of surgical patients. It is possible that the relationship between previous surgical resection and LMD may be influenced by the timing and delivery of adjuvant radiation, which is not accounted for in either analysis but may account for the difference between the 2 studies.

Table 4.

Studies Studying Patients with Melanoma Leptomeningeal Disease

	Ferguson et al.	Chorti et al	Foppen et al.	Tetu et al.	Current study
Year of diagnosis	2009–2015	2011–2019	2010–2015	2013–2020	2015–2020
Number of patients	178	52	39	29	172
Male gender	61%	58%	59%	62%	60%
Median (range) age at LMD diagnosis	51 (18–89)	58 (32–85)	52.9 (26–84)	55(50–67)	53 (20–70)
BRAF mutant	67%	65%	14 patients received BRAF/MEKi	45%	66%
% of patients with prior therapy	79%	81%	71%	52%	84%
% of patients received at least one treatment after LMD diagnosis	86%	77%	64%	93%	84%
LMD treatments	RT: N = 98 CT: N = 89 TT: N = 60 IMT: N = 12 IT: N = 64	RT: N = 3 TT: N = 17 IMT: N = 13 IT: N = 1 TT + IMT: N = 5	RT: N = 15 TT: N = 14 IMT: N = 10 IT: N = 0	RT: N = 9 CT: N = 1 TT: 5 IMT: 17 TT + IMT: N = 4	RT: N = 92 CT: N = 30 TT: N = 60 IMT: N = 57 IMT + TT: N = 30 IT: N = 42
Median OS	3.5 months	2.9 months	6.9 weeks	5.1 months	4.8 months

CT, chemotherapy; IMT, immunotherapy; IT, intrathecal therapy; RT, radiation therapy; TT, targeted therapy. LMD, leptomeningeal disease; OS, overall survival.

IT immunotherapy has been used as a treatment option for LMD, but limited to specialized treatment centers, with clinical trials ongoing (NCT05598853, NCT02271711, NCT00338377, NCT05112549). Early studies investigating the therapeutic utility of IT immunotherapy focused on interleukin-2 (IL-2).^31–34 Previously, our institution demonstrated that IT IL-2 can result in long-term survival in a subgroup of melanoma patients (n = 43) with LMD, with a median OS of 7.8 months. This result demonstrated the clinical relevance of IT immunotherapy in this patient population; however, IT IL-2 therapy has notable toxicity.³⁵ In our single institution cohort study of 178 melanoma LMD patients diagnosed from 1999 to 2015, we reported that patients treated with IT had an OS of 8 months.⁷ Similarly, in this current study, the median OS was 8 months for all patients who received IT therapy. Checkpoint inhibitors, which have higher activity and better safety, have overall replaced the systemic use of IL-2 for metastatic melanoma including patients with CNS disease and IT administration of checkpoint inhibitors is currently being evaluated in melanoma LMD. More recently, a clinical trial (NCT03025256) assessed the safety and efficacy of combined IT and IV nivolumab in melanoma patients with LMD. In an updated survival analysis with a median follow-up of 27.3 weeks (range 1–252 weeks) the authors reported an improved median OS of 7.5 months for the 50 patients treated with this regimen (48 melanoma patients, 2 NSCL patients), with OS 68%, 54%, and 35% at 3, 6, 12 months, respectively.³⁶ The interpretation of retrospective and single arm studies is naturally limited, and we acknowledge that multiple patient factors, including functional status, the status of concurrent parenchymal brain metastases and extracranial disease, prior treatment history as well as inherent physician bias will have an impact on the decision to treat LMD patients with IT therapy or immunotherapy, and patients receiving IT therapy likely have better functional status at baseline. However, we do believe that IT immunotherapy is a hopeful potential approach for this population with very limited treatment options.

In addition to IT therapy, systemic BRAF/MEK targeted therapy and checkpoint immunotherapy have both been evaluated as treatment options for melanoma LMD, albeit also either as case reports or in small series.^29,37–41 In regard to BRAF-targeted therapy, most studies and case reports have focused on vemurafenib and dabrafenib with some encouraging results. For example, in a BRAF-mutant melanoma LMD patient treated with vemurafenib followed by WBRT and a combination of dabrafenib and trametinib both survival of 19 months and radiographic response.⁴² Two additional cases reports, one using vemurafenib in a patient with poor performance, leading to significant clinical and imaging response as well as prolonged survival, while the other patient received treatment with dabrafenib and trametinib, resulting in impressive resolution of radiographic LMD and symptomatic improvement.^43,44 In regards to systemic immunotherapy, a recent Phase II study by Brastianos et al., evaluated the safety of systemic ipilimumab and nivolumab combination therapy in patients with LMD.⁹ Even though this study did not focus on melanoma (only 2 melanoma patients were included in this cohort), these authors reported that showed that 39% of patients had stable disease in the CNS as their best response and a median survival of 2.9 months. Overall, even though the results of these cases series are impressive larger prospective studies are needs to optimize this treatment strategy. Research is needed to draw significant conclusions about the most appropriate therapy for patients with LMD.

Radiation therapies for LMD had traditionally been limited to WBRT or craniospinal XRT.^22,37 As mentioned above, WBRT was not associated with improved survival in neither our prior or current study. However, there are several recent studies evaluating alternative radiation strategies for LMD. Patients with LMD traditionally received photon-based craniospinal irradiation (CSI), which is a radiotherapeutic approach that involves total or subtotal irradiation of all CNS compartments, including the whole brain and typically the complete spine. While this approach can palliate neurological symptoms it can be associated with notable toxicity. To address this, recent studies have evaluated the use of CSI using proton therapy to achieve clinical benefit with less side-effects. A recent randomized Phase II trial compared photon and proton CSI and reported that proton-based therapy had improved progression free survival and overall survival.^45,46 Only 6 few melanoma patients were included in this cohort, and further disease specific trials might be needed, but it may be a promising treatment avenue for LMD patients, particularly those with bulky and/or symptomatic disease. Additionally, with the recent acknowledgement of nodular/focal LMD, a recent study evaluated local control and overall survival of patients who underwent SRS to focal LMD (n = 16).⁴⁷ Even though this study did not focus on melanoma LMD, they reported a median post-SRS survival of 10 months and a 6-month and 1-year actuarial overall survival of 60% and 26% respectively, and also might represent a viable alternative to WBRT for patients with focal/nodular disease.

Our study has multiple inherent limitations. While this is one of the largest clinical studies dedicated to melanoma LMD, this study is limited by its retrospective nature so is subject to selection bias. Additionally, as clinical data was retrospectively collected from the electronic medical record all clinical data points were not available for every patient, which limits the robustness of the data analysis. Further, only a subset of LMD patients are candidates for multi-modality treatment approaches. Finally, patients who are considered for IT therapy, require either frequent lumbar punctures or a surgical procedure for Ommaya reservoir placement, and as such, only patients with adequate functional status are typically candidates for this therapy approach.

In conclusion, despite regulatory approval of multiple additional treatment options for metastatic melanoma including reported durable response in melanoma patients with brain parenchymal metastases, this study confirms that the overall poor prognosis of melanoma patients with LMD remain unchanged. While a subset of patients, specifically patients with high functional status, LMD-only disease and negative CSF cytology at diagnosis appear to have better outcomes, it is also important to note that >2 LMD-directed treatments were associated with longer survival. Ultimately, large prospective and ideally randomized studies are still needed to better distinguish which subset of patients benefit multi-modal therapy. Taken together, a dire need remains for increased translational studies and clinical trials dedicated to patients with melanoma LMD.

Conflict of interest statement

The authors have no conflicts of interest pertinent to this manuscript.

Funding

No funding was available for this work.

Authorship statement

C.S. collected, evaluated, and interpreted the data and co-wrote the manuscript; D.R.M. performed the statistical analysis, interpreted the results, and edited the manuscript; J.S., S.D., R.N.A., M.A.D., A.D., J.M., S.P.P., H.T., C.Y., M.K.W., and I.E.M; reviewed and revised the manuscript; S.D.F. interpreted the data and co-wrote the manuscript; I.C.G.O. collected, designed the research, analyzed and interpreted the data, and co-wrote the manuscript.

Ethics approval and consent to participate

The study protocol was approved by the Institutional Review Board of The University of Texas MD Anderson Cancer Center (IRB number PA15-0308).