Neurologic and oncologic features of Erdheim–Chester disease: a 30-patient series

Abstract

Background

Erdheim–Chester disease (ECD) is a rare histiocytic neoplasm characterized by recurrent alterations in the MAPK (mitogen-activating protein kinase) pathway. The existing literature about the neuro-oncological spectrum of ECD is limited.

Methods

We present retrospective clinical, radiographic, pathologic, molecular, and treatment data from 30 patients with ECD neurohistiocytic involvement treated at a tertiary center.

Results

Median age was 52 years (range, 7–77), and 20 (67%) patients were male. Presenting symptoms included ataxia in 19 patients (63%), dysarthria in 14 (47%), diabetes insipidus in 12 (40%), cognitive impairment in 10 (33%), and bulbar affect in 9 (30%). Neurosurgical biopsy specimens in 8 patients demonstrated varied morphologic findings often uncharacteristic of typical ECD lesions. Molecular analysis revealed mutations in BRAF (18 patients), MAP2K1 (5), RAS isoforms (2), and 2 fusions involving BRAF and ALK. Conventional therapies (corticosteroids, immunosuppressants, interferon-alpha [IFN-α], cytotoxic chemotherapy) led to partial radiographic response in 8/40 patients (20%) by MRI with no complete responses, partial metabolic response in 4/16 (25%), and complete metabolic response in 1/16 (6%) by ¹⁸F-fluorodeoxyglucose (FDG)-PET scan. In comparison, targeted (kinase inhibitor) therapies yielded partial radiographic response in 10/27 (37%) and complete radiographic response in 14/27 (52%) by MRI, and partial metabolic response in 6/25 (24%) and complete metabolic response in 17/25 (68%) by FDG-PET scan.

Conclusions

These data highlight underrecognized symptomatology, heterogeneous neuropathology, and robust responses to targeted therapies across the mutational spectrum in ECD patients with neurological involvement, particularly when conventional therapies have failed.

Key Points.

1. This comprehensive overview is the largest neuro-oncological series of patients with ECD to date.

2. There are underrecognized symptomatology and heterogeneous neuropathology with robust responses to targeted therapies in central nervous system histiocytoses.

Importance of the Study.

This represents to our knowledge the largest neuro-oncologic series of patients with ECD to date, encompassing clinical, radiologic, pathologic, mutational, and treatment data. Our data suggest dense and underappreciated neurologic symptomatology in ECD patients, yet robust and consistent responses to targeted therapies.

Erdheim–Chester disease (ECD) is a rare, non-Langerhans cell histiocytosis characterized by multi-organ infiltration of histiocytes, chronic inflammation, and fibrosis. ECD represents a heterogeneous clinical spectrum ranging from limited to disseminated, life-threatening forms. ECD was initially considered an immune granulomatous disease; however, identification of recurrent activating mutations in the mitogen-activating protein kinase (MAPK) pathway^1–4 in lesional tissue has led to the redefinition of ECD as a clonal hematopoietic disorder of the monocyte/macrophage and dendritic cell lineages.^5,6 Targeted therapies have emerged as new therapeutic strategies for ECD with robust efficacy reported with BRAF and MEK inhibitors.^7–10

Tumorous infiltration in ECD commonly involves the bones, skin, retroperitoneum, heart, orbits, lungs, and brain.^11–13 Involvement of the nervous system over the ECD disease trajectory is estimated to occur in up to one third of cases, and neurologic involvement was independently associated with death in a large series.¹⁴ The existing literature about the manifestations, imaging, pathology, and treatment of neurologic ECD is limited predominantly to single case reports or small series,^15–22 with the exception of one larger series dedicated to neuroimaging^22,23 and histopathology.^24,25 There exist little data collected systematically from large cohorts about neurologic manifestations, imaging findings, histopathology, and treatment outcomes in ECD. Furthermore, there has been little examination of the spectrum of activating mutations present in BRAFV600-wildtype neurologic ECD patients. We present here comprehensive data from 30 patients with ECD neurohistiocytic involvement. Our purpose is to provide clinicians across neurologic disciplines with the largest available dataset of the neuro-oncologic spectrum of ECD to aid in the evaluation and management of the ECD patient.

Materials and Methods

Patient Selection

This retrospective study was approved with a waiver of informed consent by the institutional review board. Included were patients with ECD or mixed ECD/Langerhans cell histiocytosis (ECD/LCH) or mixed ECD/Rosai–Dorfman disease (ECD/RDD), with involvement of cranial or spinal structures at any time in the disease trajectory, evaluated at Memorial Sloan Kettering Cancer Center between January 1, 2013 and February 1, 2019. All patients were diagnosed with ECD based on published criteria²⁶ characterized by distinctive histiocyte infiltration with associated inflammation on tissue biopsy (of the nervous system or other sites) with concordant radiographic findings of symmetric osteosclerosis in the legs on bone scintigraphy or ¹⁸F-fluorodeoxyglucose positron emission tomography (FDG-PET), in the setting of characteristic systemic findings.

Clinical and Neuroimaging Data Review

Medical records were independently assessed by two board-certified neuro-oncologists (A.B. and E.L.D.), including demographic characteristics, neurologic and systemic symptoms, biopsy sites, duration of symptoms prior to diagnosis, and treatment outcomes. Treatments received were categorized into surgical resection, radiation therapy, conventional systemic therapy (corticosteroids, immunosuppressants, interferon-alpha [IFN-α], cytotoxic chemotherapy), and targeted kinase inhibitors (eg, BRAF, MEK inhibitors).

Clinical Assessments

Clinical responses to treatment, as documented in the medical record, were subjectively classified as (i) complete resolution (CR) of symptoms attributed to ECD, (ii) partial resolution (PR) of symptoms—without a thresholded reduction in lesional size, (iii) stable disease (SD) symptoms, or progressive disease (PD) symptoms, as previously published.²⁷

Neuroimaging Assessments

Contrast-enhanced MRI of the head was performed for all patients and reviewed by a board-certified neuroradiologist (V.H.) experienced in the assessment of ECD neuroimaging. The presence or absence of disease was documented involving the orbits, paranasal sinuses, calvarium, skull base (osseous and extraosseous components), dura, leptomeninges, supratentorial and infratentorial brain parenchyma, and hypothalamic-pituitary axis. Overall parenchymal brain volume of the supratentorial and infratentorial compartments was inspected for the presence of generalized atrophy. Additionally, perfusion characteristics of intracranial abnormalities were assessed when dynamic contrast-enhanced (DCE) MR imaging was performed. Blood plasma volume (Vp) and volume transfer constant (K^trans) of the intracranial ECD lesions were categorized as elevated or not elevated with respect to normal brain parenchyma. MR images of the entire spine, when performed, were analyzed for the presence of osseous, epidural, dural-based, leptomeningeal, and intramedullary lesions. Responses to treatment as measured by MRI were determined from clinical neuroradiologic interpretation of scans at our institution or, when scans were not reviewed at our institution, from outside medical records. Responses were categorized, as has been published in ECD²⁷ as CR, PR, SD, or PD (growth of known lesions or appearance of new lesions).

Lastly, as FDG-PET is considered the optimal response assessment for ECD and has been implemented as an endpoint in clinical trials of ECD therapies,^8,9 instances when FDG-PET was performed prior to starting treatment and to assess treatment response were reviewed. Metabolic responses by FDG-PET were characterized as (i) CMR (complete metabolic resolution of FDG-avid lesions), (ii) PMR (partial metabolic resolution of FDG-avid lesions—without a thresholded reduction in lesional FDG uptake), (iii) SD (stable lesional avidity), or (iv) PMD (progressive metabolic disease).

Neuropathologic and Molecular Analysis

Tumor material obtained from neurosurgical biopsies was reviewed by an experienced neuropathologist (M.K.R.). When tissue was available for sequencing (from neurologic or systemic samples), genomic DNA was extracted from fresh-frozen or formalin-fixed paraffin-embedded samples and underwent mutational analysis using previously described techniques, including whole-exome sequencing, allele-specific polymerase chain reaction (PCR) for BRAFV600E, or Memorial Sloan Kettering–Integrated Mutation Profiling of Actionable Cancer Targets, a targeted hybrid capture next-generation sequencing–based DNA sequencing panel of 400 cancer genes with a matched-normal sample to exclude germline variants.^1,28,29 For some patients, a next-generation sequencing–based RNA-fusion detection assay (Archer FusionPlex Custom Solid Panel) was implemented to detect fusions across 62 cancer-related genes involved in chromosomal rearrangements.³⁰ One patient underwent autopsy.

Statistical Analysis

Clinical, radiologic, molecular, and treatment response data were analyzed using descriptive statistics. Continuous variables were summarized with means and ranges, and categorical variables were summarized with frequencies and proportions. Conventional systemic therapies were descriptively compared with targeted therapies with respect to frequency of clinical and imaging responses.

Results

Presenting Symptoms and Sites of ECD Involvement

Thirty patients with ECD were included (Table 1). The median age at ECD diagnosis was 52 years (range, 7–77), and 20 (67%) patients were male. Six patients (20%) had mixed histiocytosis. The median duration of symptoms prior to ECD diagnosis was 21.5 months (range, 1–528 mo). The most common presenting neurological symptoms were ataxia in 19 (63%) patients and dysarthria in 14 (47%) patients. Diabetes insipidus occurred in 12 (40%) patients. Cognitive impairment and brainstem symptoms such as diplopia were each reported in 10 (33%) patients. Bulbar affect occurred in 9 (30%) patients followed by hemiparesis in 8 (27%), and dysphagia in 7 (23%). Fatigue in 24 (80%) patients and bone pain in 15 (50%) patients were the two most common systemic symptoms.

Table 1.

Clinical characteristics and presenting symptoms in patients with ECD involvement of the nervous system (N = 30)

Characteristic	No / Median	% / (Range)
Male	20	67
Female	10	33
Age, y, at ECD diagnosis	52	(7–77)
Duration, mo, of neurologic symptoms before ECD diagnosis	21.5	(1–528*)
Status at last follow-up
Alive	28	93
Dead	2	7
Mixed histiocytosis	6	20
ECD/LCH	4	13
ECD/RDD	2	6
Presenting signs and symptoms
Neurologic
Ataxia	19	63
Dysarthria	14	47
Diabetes insipidus	12	40
Cognitive impairment	10	33
Diplopia	10	33
Bulbar affect	9	30
Hemiparesis	8	27
Dysphagia	7	23
Headache	5	17
Proptosis	4	13
Hearing loss or tinnitus	3	10
Seizures	2	6
Urinary symptoms	2	6
Hemisensory deficit	1	3
Systemic
Fatigue	24	80
Bone pain	15	50
Mood changes	10	30
Arthralgias/myalgia	7	23
Night sweats	7	23
Hypogonadism	7	23
Abdominal pain	6	20
Dyspnea	5	17
Skin lesions	4	13
Weight loss	4	13
Back pain	4	13
Xanthalesma	3	10
Fever	2	7
Amenorrhea	2	7
Chills	1	3
ECD biopsy site** (N = 52 biopsies in 30 patients)
Bone	15	29
Brain	8	15
Skin	5	10
Retroperitoneum	5	10
Testicle	3	6
Orbit	3	6
Lung	3	6
Muscle/soft tissues	3	6
Gastrointestinal	2	13
Dura	1	2
Spine	1	2
Liver	1	2
Sinus	1	2
Tongue	1	2
ECD Mutation
BRAFV600E	18	60
MAP2K1	5	17
NRAS	1	3
KRAS	1	3
BRAF-PICALM fusion	1	3
KI5FB-ALK fusion	1	3
No mutation identified	2	7
Sequencing not performed	1	3

*This patient presented with diabetes insipidus at age 4 and was diagnosed with ECD at age 48.

**Several patients underwent more than one biopsy procedure. There were 52 total biopsies among the 30 patients.

Neuroimaging

Contrast-enhanced MR images of the head were analyzed for 30 patients (Table 2). In 19 patients, the MRI analyzed was performed at the time of ECD diagnosis and in 11 the MRI was performed subsequent to diagnosis. Findings are below:

Table 2.

Radiologic involvement in patients with ECD of the nervous system (N = 30)

Cranial bones	n	(%)
Calvarium	13	43
Skull base	8	27
Mastoid	1	3
Extraosseous skull base	5	17
Paranasal sinuses	8	27
Orbits	5	17
Hypothalamic-pituitary axis	13	43
Dura (cranial)	10	33
Brain
Parenchyma	18	60
Supratentorial	11	37
Infratentorial	16	53
Leptomeninges	0	0
Cerebellar atrophy	4	13
Spine
Vertebrae	4	57
Epidural space	1	14
Dura	0	0
Intramedullary spinal cord	1	14
Leptomeninges1	1	14
Systemic disease involvement
Bones	30	100
Retroperitoneum	18	60
Abdomen (omentum or mesentery)	10	33
Peri-aortic	8	27
Lungs or pleura	8	27
Cardiac	7	23
Testes	5	17
Skin	4	13
Lymph nodes	4	13
Thyroid	2	7
Subcutaneous soft tissue	2	7
Rectal mass	2	7
Liver	1	3

¹Confirmed by cerebrospinal fluid analysis.

Osseous structures, orbital structures, and hypothalamic-pituitary axis (HPA)

(Figure 1)Fifteen (50%) patients had at least one site of osseous disease. The calvarium was the most common site of cranial bone involvement, occurring in 13/30 (43%) patients, followed by the osseous skullbase (27%) and paranasal sinuses (27%). Three patients with extraosseous skull base lesions had concomitant infiltration of the adjacent bones and 2 patients had extraosseous skullbase disease without evidence of bone involvement. Orbital lesions were identified in 5 (17%) patients and demonstrated contrast enhancement. The intra- and extraconal compartments were both involved in 3 patients. The remaining 2 patients had isolated intraconal and extraconal orbital infiltration. Thirteen (43%) patients had lesions along the hypothalamic-pituitary axis.

Fig. 1.

Neuroradiologic ECD. (A) Irregular enhancing lesions in the middle cerebellar peduncles are demonstrated by post-gadolinium axial T1 MRI. (B) Supratentorial and infratentorial enhancing ECD lesions and (C) enhancing hypothalamic lesion and (D) dural ECD lesion in the middle temporal fossa on post-gadolinium coronal T1 MRI. Sagittal fused FDG-PET/CT images demonstrate hypermetabolic epidural spinal ECD, and (F) axial post-gadolinium T1 MRI show leptomeningeal ECD infiltration. (G) ECD involvement of the optic nerves with extension to the cavernous sinus (arrowheads) is shown on post-gadolinium axial T1 MRI. (H) Atrophy of the cerebellum is demonstrated by sagittal T1 MRI.

Brain parenchyma

Parenchymal lesions in the supratentorial compartment and/or posterior fossa were present in 18 (60%) patients and were always enhancing, although variably with respect to degree/avidity and homogeneity of contrast enhancement. Parenchymal lesions demonstrated both nodular/well-defined and infiltrative enhancement patterns, even within the same patient. Similarly, the T1- and T2-weighted noncontrast imaging characteristics of the parenchymal lesions were heterogeneous.

Other structures and findings

Dural-based lesions were present in 10 (33%) patients and were invariably contrast enhancing. Dural lesions were T2-hypointense relative to the brain parenchyma in 6 patients and iso- to hyperintense in 1 patient. Seven patients had DCE perfusion sequences performed. The parenchymal and/or dural lesions demonstrated elevated K^trans without Vp elevation in 5 patients and elevation of both parameters in 1 patient. The lesions for the other patient were too small to adequately characterize on perfusion. Four patients (13%) had overt cerebellar atrophy out of proportion to volume loss in cerebral hemispheres. There were no instances of intracranial leptomeningeal disease. Dedicated MRI of the spine was performed in 7/30 patients and demonstrated vertebral lesions in 4 (57%), epidural lesions in 1 (14%), no instances of dural ECD, intramedullary disease in 1 (14%), and overt leptomeningeal involvement in 1 (14%).

Neuropathologic Evaluation and Molecular Diagnostics

ECD assumed varied histologic guises in the 8 patients with neurosurgical biopsies (Figure 2). Extra-axial lesions involving cranial bone or dura were most likely to harbor the signature foamy histiocytes in conspicuous numbers, these associated with various degrees of lymphoplasmacellular infiltration and fibroplasia. Neuroparenchymal lesions in select cases also evidenced obvious perivascular and interstitial foam cell infiltration, but this tended to be less conspicuous and was altogether absent from some examples. Foci of neuroparenchymal ECD often presented with the nonspecific chronic inflammatory picture of perivascular lymphocytic cuffing (mainly by small T cells, B cells being present in lesser numbers), astrogliosis (with prominent Rosenthal fiber formation in some instances), and tissue hypercellularity reflecting infiltration by non-foamy histiocytes that were only identifiable as such owing to their immunohistochemical expression of CD68 or CD163. Some examples also manifested tissue rarefaction with patchy loss of Luxol fast blue/myelin staining and relative axonal preservation suggesting demyelinating disease, but such lesions never demonstrated the sharp circumscription and total myelin loss with axonal sparing characteristic of primary demyelination. In all cases, the histiocytes present failed to label for CD1a, although S-100 reactivity was encountered in some neuroparenchymal histiocytic infiltrates.

Fig. 2.

Heterogeneous neuropathology of ECD. (A) Infiltrate of foamy histiocytes (arrow) with associated fibrosis characterizes this prototypical ECD lesion (hematoxylin-eosin; 200x magnification; scale bar = 100 µm). (B) Immunohistochemical assessment demonstrates CD68 (a macrophage marker) expression by histiocytes (hematoxylin counterstain; 200x magnification; scale bar = 100 µm). (C) ECD presenting with a nonspecific, chronic inflammatory/reactive picture. Note modest cuffing of blood vessel at upper left by small lymphocytes. Mononuclear cells are present in the neuroparenchyma but do not exhibit foam cell changes and are not clearly recognizable as histiocytes. A Rosenthal fiber is present to right of center (hematoxylin-eosin; 200x magnification; scale bar = 100 µm). (D) A second field from the same case demonstrates prominent Rosenthal fiber formation (arrows), a reactive phenomenon that prompted diagnostic consideration of adult-onset Alexander’s disease (hematoxylin-eosin; 200x magnification; scale bar = 100 µm). (E) Autopsy section of the infundibular stalk/neurohypophysis demonstrates a nonspecific increase in mononuclear cells (hematoxylin-eosin; 200x magnification; scale bar = 100 µm). (F) Immunohistochemical assessment shows the cells in the prior figure to include CD163-expressing (another macrophage marker) histiocytes (hematoxylin counterstain; 200x magnification; scale bar = 100 µm). (G) Nonspecific, chronic inflammatory presentation of ECD. Note that small lymphocytes dominate the field in the absence of foam cells (hematoxylin-eosin; 200x magnification; scale bar = 100 µm). (H) Immunohistochemical assessment of the same case demonstrates the presence of many CD163-expressing histiocytes (hematoxylin counterstain; 200x magnification; scale bar = 100 µm).

One patient underwent autopsy following ECD-related death (Figure 2). MRI had previously revealed no evidence of ECD involvement and the patient did not have diabetes insipidus. Gross examination of the nervous system was normal; however, microscopic examination of pituitary and infundibular sections demonstrated conspicuous infiltration of parapituitary epidural tissue by foamy histiocytes and small lymphocytes, consistent with involvement of ECD. The leptomeninges sheathing the infundibular stalk and the infundibulum demontrated histiocytic infiltration, also consistent with ECD.

Molecular analysis demonstrated a somatic alteration in 27 (90%) patients (Figure 3): 18 (60%) patients were found to have the BRAFV600E mutation, 5 (17%) had MAP2K1 mutations, 2 (7%) carried RAS mutations, 1 (3%) had a BRAF-PICALM fusion mutation, and 1 (3%) had a KIF5B-ALK fusion. Two patients (7%) did not have a mutation identified by sequencing performed on tumor material, and 1 patient (3%) did not have sufficient material for molecular analysis.

Fig. 3.

Molecular landscape of kinase alterations in neurologic Erdheim–Chester disease. (A) Oncoprint of kinase alterations with each patient represented in one column. Histiocytosis diagnoses (primary and secondary, if applicable) and sequencing methodology are in the first 3 rows. Somatic driver kinase alterations are shown in the lower rows. (B) Pie chart of kinase alterations in these 30 patients. Protein diagrams of (C) somatic kinase mutations and (D) protein fusions.

Treatments

Twenty-nine of 30 (97%) patients were treated for ECD with neurologic involvement (Table 3, Supplementary Table 2); 1 patient was diagnosed with neurologic ECD at autopsy and was therefore not treated for this disease manifestation. Zero patients underwent surgical resection. Three patients underwent radiotherapy; 1 received whole-brain radiotherapy (WBRT) with a boost to the posterior fossa; 1 received intensity-modulated radiotherapy (IMRT) to the brainstem; and another received stereotactic radiosurgery. The patient who received WBRT had a partial clinical response and CT/MRI response, and the other 2 had progressive disease.

Table 3.

Conventional and targeted therapies in ECD of the nervous system (N = 30)

	ECD Localization in Cranial and Spinal Structures	Conventional Treatment	Clinical	CT/MRI	PET	Mutation	Targeted Treatment	Clinical	CT/MRIa	PETb	Duration of Response to Targeted Therapy	Relapse on Targeted Therapy	Vital Status
1	Brain parenchyma, supratentorium, infratentorium, spinal cord, vertebrae	Corticosteroids	PD	PD	PMD	BRAF V600E	Vemurafenib, Dabrafenibd	PR	CR	CMR	>5 years	No	Alive
		HD-MTX	PD	PD	PMD
		Interferon-alpha	ID	ID	ID
2	HPA, disproportionate atrophy	–	–	–	–	BRAF V600E	Vemurafenib, Dabrafenib, Trametinibc	PR	PR	CMR	>3 years	No	Alive
3	Dura, orbits, extraosseous skull base	–	–	–	–	BRAF V600E	Vemurafenib, Dabrafenibc	PR	PR	PMR	>5 years	No	Alive
4	Dura, HPA, orbits, skull base, paranasal sinuses	–	–	–	–	BRAF V600E	Vemurafenib	CR	PR	CMR	>5 years	No	Alive
5	Brain parenchyma, infratentorium	HD-MTX	PR	PR	ID	BRAF V600E	Vemurafenib	PR	CR	CMR	>5 years	No	Alive
6	Cerebellar atrophy	Interferon-alpha	PD	PD	ID	BRAF V600E	–	–	–	–	–	–	Alive
7	Brain parenchyma, supratentorium, infratentorium, HPA, orbits, calvarium, extraosseous skull base, paranasal sinuses	Corticosteroids	PD	PD	ID	BRAF V600E	Vemurafenib, Dabrafenibc	CR	CR	CMR	>3 years	No	Alive
8	Brain parenchyma, infratentorium, HPA, calvarium, skull base, paranasal sinuses	Cytarabine	SD	PR	SMD	BRAF V600E	Dabrafenib	PR	CR	CMR	>3 years	No	Alive
9	Brain parenchyma, infratentorium, dura, HPA, calvarium, skull base	Interferon-alpha	PD	ID	ID	BRAF V600E	Imatinib	PD	PD	ID	>3 years	No	Alive
		Vinblastine/ prednisone	PD	ID	ID
		6-mercaptopurine	PD	ID	ID		Vemurafenib, Cobimetinibc	SD	PR	CMR
		Infliximab	PD	PD	ID
		Methotrexate	PD	PD	ID
10	HPA, vertebrae	Corticosteroids	SD	SD	SMD	BRAF V600E	Cobimetinib	PR	PR	CMR	>3 years	No	Alive
11	Brain parenchyma, supratentorium, infratentorium, HPA, dura, calvarium	–	–	–	–	BRAF V600E	Vemurafenib, Dabrafenibc	PR	CR	PMR	>5 years	No	Alive
12	Brain parenchyma, infratentorium	Corticosteroids	PD	PD	ID	BRAF V600E	Dabrafenib	PR	CR	CMR	>2 years	No	Alive
13	Brain parenchyma, supratentorium, infratentorium, dura, calvarium, skull base, paranasal sinuses	–	–	–	–	BRAF V600E	Cobimetinib	CR	CR	CMR	>3 years	No	Alive
14	Brain parenchyma, supratentorium, infratentorium, dura, calvarium, skull base, vertebrae	–	–	–	–	BRAF V600E	Dabrafenib	PR	PR	PMR	>2 years	No	Alive
15j	Dura, calvarium, paranasal sinuses	–	–	–	–	BRAF V600E	Dabrafenib	CR	CR	CMR	>2 years	No	Alive
16	Brain parenchyma, supratentorium, cerebellar atrophy, HPA, dura, calvarium, skull base, extraosseous skull base, paranasal sinuses,	Corticosteroids	PD	PD	ID	BRAF V600E	Cobimetinib	PR	PR	PMR	>2 years	No	Alive
		IVIG	PD	PD	ID
		Rituximab	PD	PD	ID
		Cyclophosphamide	PD	PD	ID
		HD-MTX	PD	PD	ID
		Lenalidomide +  etoposide +  dexamethasone	SD	SD	SMD
17	Brain parenchyma, infratentorium, HPA, calvarium	Vinblastine	PD	PD	ID	BRAF-PICALM fusion	Cobimetinib	PR	CR	IDf	>2 years	No	Alive
		Etoposide	PD	PD	ID
		Interferon	PD	PD	ID
18	Orbits, brain parenchyma, supratentorium, infratentorium, HPA, calvarium	Anakinra	PR	PR	PMR	MAP2K1 C121S	Trametinib	SD	PR	PMR	6 months	No	Deadh
19	Epidural (spine)	Cellcept	PD	PD	ID	MAP2K1 E51G	–	–	–	–	–	–	Alive
		Corticosteroids	PD	PD	ID
		Anakinra	PR	PR	PMR
20	Calvarium, skull base, extraosseous skull base, paranasal sinuses, dura	Anakinra	SD	SD	SMD	MAP2K1 P124L	Cobimetinib	PR	SD	CMR	9 months	No	Deadi
21	HPA, extraosseous skull base	Corticosteroids	ID	ID	ID	MAP2K1 P105I107indel	Cobimetinib	CR	CR	CMR	>2 years	No	Alive
		Anakinra	ID	PD	PMD
22	Brain parenchyma, supratentorium, infratentorium, cerebellar atrophy	Interferon-alpha	PD	PD	PMD	KRAS R149G	Dasatinibd	SD	SD	SMD	>3 years	No	Alive
		Sirolimusd	SD	SD	SMD		Cobimetinib	PR	PR	PMR
23	Orbits, dura	Corticosteroids	PR	PR	PMR	NRAS Q61R	Trametinib	PR	CR	CMR	>5 years	No	Alive
		Interferon-alpha	PR	PR	PMR
24	Leptomeningeal disease, calvarium, skull base, vertebrae	Interferon-alpha	PD	PD	PMD	KIF5B-ALK fusion	Crizotinib	PR	PR	CMR	>2 years	No	Alive
25	Brain parenchyma, supratentorium	Corticosteroidse	CR	PR	CMR	no mutation identified	–	–	–	–	–	–	Alive
26	Brain parenchyma, supratentorium, infratentorium, HPA, calvarium	Corticosteroids	PD	PD	ID	no mutation identified	–	–	–	–	–	–	Deadj
		IVIG	PD	PD	ID
		Cellcept	PD	PD	ID
		HD-MTX	PD	PR	ID
27	Brain parenchyma, supratentorium, infratentorium, HPA	Vinblastine/prednisone	PD	PD	PMD	no mutation identified	Trametinib	CR	CR	CMR	>3 years		Alive
		Anakinra	SD	PD	ID
28	Brain parenchyma, supratentorium, infratentorium	Corticosteroids	SD	SD	ID	BRAF V600E	Dabrafenib +  trametinib	PR	CR	CMR	1 year	No	Alive
29	Brain parenchyma, Infratentorium, HPA	Corticosteroids	PD	PD	ID	BRAF V600E	Dabrafenib +  trametinib	CR	CR	–	>3 years	No	Alive
		Interferon-alpha	SD	SD	ID
30g	HPA	–	–	–	–	MAP2K1 K57N	–	–	–	–	1 year	No	Deadk

Abbreviations: PD, progressive disease; SD, stable disease; PR, partial response; CR, complete response; PMD, progressive metabolic disease; SMD, stable metabolic disease; PMR, partial metabolic response; CMR, complete metabolic response; ID, insufficient data: response assessment modality not performed and/or not documented.

^aMRI response to all lesions in nervous system (brain and spine).

^bFDG response to all lesion in nervous system (brain and spine).

^cVemurafenib discontinued for intolerance or toxicity, in the setting of favorable response, in favor of dabrafenib, trametinib, or cobimetinib.

^dDasatinib was given in combination with sirolimus for this patient.

^eUnusual response to steroid monotherapy.

^fLesion not measurable by FDG-PET which was performed.

^gDural disease identified postmortem, treatment response not considered.

^hDied of ischemic cardiac disease. Response to trametinib had been sustained.

ⁱDied of cardiac failure in the setting of sepsis. Response to cobimetinib had been sustained.

^jDied of neurologic progression, prior to the era of treatment with targeted therapies.

^kDied of peritonitis. Response to trametinib (systemic) had been sustained.

Of the 29 patients treated for neurologic ECD, 4 (14%) received conventional therapy only, 7 (24%) received targeted therapy only, and 18 (62%) received conventional therapy followed by targeted therapy. Of all (n = 22) patients receiving conventional therapies (Table 3), 12 (55%) received steroids, 8 (36%) received IFN-α, 9 (41%) received cytotoxic chemotherapy (cytarabine, cladribine, cyclophosphamide, vinblastine, etoposide, lenalidomide), 9 (41%) immunomodulatory therapies (anakinra, Cellcept, intravenous immunoglobulin [IVIG]) during their course of treatment. Eleven (38%) patients receiving conventional therapy received 2 or more of these agents. Viewed in aggregate, there were 45 instances of conventional therapy across all 29 patients. Of the 25 patients receiving targeted therapies, 12 (48%) received BRAF inhibitor, 11 (44%) MEK inhibitor, and 2 (8%) combined BRAF/MEK inhibitors. Of these patients, 1 received imatinib and 1 dasatinib prior to BRAF or MEK inhibitors. In aggregate, there were 27 instances of targeted therapy in our cohort.

Clinical responses

(Table 3)Considering 42 instances of conventional therapy with sufficient data among the patients to designate a subjective clinical response, 28 (67%) led to progressive clinical symptoms, 8 (19%) to stable symptoms, and 6 (14%) to partial or complete resolution of symptoms. Where sufficient clinical response data existed, complete or partial clinical responses were seen in 2/11 (18%) instances of corticosteroids, 2/11 (18%) instances of immunosuppressive agents, 1/7 (14%) instance of IFN-α, and 1/13 (8%) instance of chemotherapy. By comparison, out of the 27 instances of targeted therapies, there was 1 (4%) subjective progression of clinical symptoms, 3 (11%) stable symptoms, and 23 (85%) partial or complete resolution of symptoms. Complete or partial clinical responses were seen in 11 (100%) of those taking BRAF inhibitors, 9/10 (90%) of those receiving MEK inhibitors, and 2/3 (67%) of those on combined BRAF/MEK therapy.

MRI responses

Response assessment in the head or spine by MRI was documented for 40 (89%) of the instances of conventional therapy and 27 (100%) of targeted therapy (Table 3). Overall conventional therapies conferred no complete responses and a partial response in 8 (20%) instances, stable disease in 6 (15%) instances, and progression of disease in 26 (65%) instances. Where sufficient MRI response data existed, complete or partial responses were seen in 3/11 (27%) instances of chemotherapy, 2/12 (17%) instances of immunosuppressive agents, 1/7 (14%) instance of IFN-α, and 2/11 (18%) instances of corticosteroids. Alternatively, targeted therapy led to a partial MRI response in 10/27 (37%) or complete MRI response in 14/27 (52%) instances, stable MRI response in 2/27 (7%) instances, and progressive disease in 1/27 (4%) instance.

FDG-PET response

Response assessment by FDG/PET was documented for 16 (36%) of the instances of conventional therapy and 25 (93%) of targeted therapy (Table 3). Overall conventional therapies conferred partial metabolic response in 4/16 (25%) and complete metabolic response in 1/16 (6%) instances, stable metabolic disease in 5/16 (31%) instances, and progressive metabolic disease in 6/16 (38%) instances. Where sufficient FDG-PET response data existed, complete or partial responses were seen in 2/4 (50%) instances of corticosteroids, 2/5 (40%) instances of immunosuppressive agents, 1/3 (33%) instance of IFN-α, and zero instances of chemotherapy. In contrast, 6/25 (24%) instances of targeted therapies had a partial metabolic response and 17/25 (68%) had a complete metabolic response, 1 (4%) had stable metabolic response, and zero had progressive disease.

Duration of responses and survival

Responses to targeted BRAF or MEK therapy were durable (see Supplementary Table 2 for duration of response, per patient, and current ECD therapy), and there were zero instances of progressive disease in the setting of ongoing BRAF or MEK inhibitor therapy. Of the 30 patients in this series, 26 are alive and 4 are deceased. One of these died of progressive neurologic ECD, prior to the era of targeted therapies, and 3 died of systemic processes not directly related to ECD.

Discussion

In this series, we present 30 patients with neurologic ECD involvement, including symptomatology, neuroimaging, neuropathology, mutational status, and treatment outcomes. This cohort demonstrated varied neurologic symptoms with notable frequency of diabetes insipidus, symptoms related to the posterior fossa, as well as cognitive impairment. The most frequently involved structures by MRI were the bones and supratentorial and infratentorial brain parenchyma. Brain biopsies demonstrated heterogeneous morphologic findings rarely consistent with classic ECD histopathology by virtue of absent xanthogranulomatous histiocytic infiltration in nearly all cases. Conventional therapies were less effective than targeted therapies in treating neurologic ECD in the patients reviewed here.

The neurologic symptomatology in this cohort bears similarities to, but important differences from, that which has been described previously in this clinical context. We observed a long duration of symptoms prior to ECD diagnosis (median 21.5 mo), as has been observed in other cohorts,³¹ which likely reflects the insidious onset of ECD in many cases, as well as the diagnostic delay characteristic of rare and unrecognized diseases. In addition to commonly reported neurological symptoms in ECD,^{13,25,32–35} we observed other symptoms that have been infrequently reported, such as bulbar affect and cognitive impairment in 30% and 33% of the cohort, respectively, as well as headache, hearing loss, and tinnitus. Conversely, we did not observe symptoms such as trigeminal neuralgia, homonomous hemianopia, and movement disorders that have been observed by others.^36–39 These findings emphasize to the clinician the spectrum of potentially underappreciated neurologic symptoms referable to ECD, in particular those more likely to be elicited by a neurologically oriented assessment such as cognitive impairment, and furthermore some that are potentially treatable (eg, bulbar affect). The issue of cognitive impairment remains a pressing and poorly understood ECD phenomenon; a recent study of patient-reported outcomes in 50 patients enrolled in a prospective ECD registry study, with and without brain involvement, demonstrated that 52% endorse symptoms of impaired attention and memory. Additionally, a retrospective volumetric neuroimaging study of 11 ECD patients without supratentorial CNS lesions demonstrated reduction of cerebral gray matter volumes in ECD patients compared with matched controls.¹⁹ These phenomena remain unclear and are the subject of a prospective study (NCT03127709).

Similar to the clinical findings in this series, the neuroradiologic features observed here both corroborate and challenge the existing knowledge about neurologic ECD. Our findings are similar to those of several case reports and two larger case series which demonstrate predominant involvement of the cranial bones, sinuses, orbits, hypothalamic-pituitary axis, dura, both supra- and infratentorial parenchymal structures, and spinal column.^{15,17,21–23,40–44} Our series suggests that dedicated CT of the sinuses and/or skullbase may be helpful to identify culprit osseous lesions in patients with potentially referable neurologic symptomatology. We observed ECD within the spinal cord only once, consistent with the rarity with which this has been described.^20,45,46 The relative infrequence of clinically significant spine lesions argues against screening spine imaging in the absence of localizing symptoms or findings. Additionally, we describe a unique case of leptomeningeal dissemination of ECD. Four patients had overt cerebellar atrophy in the absence of tumorous lesions, a finding yet to be reported as a recurrent phenomenon in ECD. This finding is notably similar to neurodegeneration observed in association with pediatric LCH, itself an enigmatic entity with respect to its pathogenesis. Among many other questions, the potential shared mechanism of these entities in ECD and LCH is of great interest to the field of neurohistiocytosis and in need of dedicated investigation.

The protean neuropathologic findings in this series are instructive with respect to the interpretation of neurosurgical biopsies and the importance of suspecting ECD in the absence of typical ECD histopathology. In these patients, the presence of histiocytes in parenchymal biopsies was scant, or inapparent in the absence of CD68 or CD163 immunostaining, with predominant findings of chronic inflammation. Rosenthal fibers were observed, as has been published previously in another case,⁴⁷ as well as findings reminiscent of demyelinating disease. These results emphasize the importance of maintaining suspicion for ECD, in the correct clinical context, even in the absence of typical findings upon brain biopsy. Conversely, these results suggest that ECD should be considered in the differential diagnosis when a brain biopsy appears to raise concern for non-neoplastic inflammatory or demyelinating disease, again in the appropriate context. Our identification of several non-BRAFV600E alterations in these patients extends the spectrum of mutations associated with neurologic ECD.

Within the limits of these data, our study suggests that targeted therapies confer marked clinical and radiologic responses in neurologic ECD. Our finding of predominantly favorable responses to targeted therapies is consistent with recent prospective trials of vemurafenib and cobimetinib for histiocytosis. Furthermore, treatment with targeted BRAF or MEK was remarkably durable (6 mo to greater than 5 y) with zero instances of relapsed disease in the setting of ongoing targeted therapy. We interpret the relatively poor responses to conventional therapies in these patients, however, with some caution. First, the response assessments performed among patients treated with conventional therapies were less comprehensive in that they rarely involved FDG-PET, which has emerged as the primary imaging modality for evaluating ECD therapy. This is somewhat overshadowed, however, by the significant proportion (64%) of progressive clinical disease among those treated with conventional therapies. Second, the patient population at our center is partially selected for patients with ECD that has been unresponsive to conventional therapies. Lastly, the patients in this series did not receive the two conventional therapies with the most ample evidence for activity in CNS ECD, namely prolonged high-dose INF-α ⁴⁸ and cladribine.²⁷ We note that favorable responses were seen in 2 of 3 patients with intravenous high-dose methotrexate (HD-MTX), which has been observed in one case report.¹⁶ Altogether, we would propose the inference from our data that targeted therapies appear to be highly effective, and moreover effective in instances of failure of conventional therapies. Even if the nature of these data limits the conclusions to be drawn about the effectiveness of conventional therapies generally, our current practice is to initiate targeted therapy as first-line treatment (BRAF inhibitor therapy for BRAFV600E-mutated disease, MEK inhibitor therapy for BRAFV600E-wildtype disease) for patients with neurologic ECD involvement, especially in the setting of symptomatic or parenchymal disease. Optimal dosing and duration of treatment remains unknown, although the patients in this series largely remained on indefinite therapy.

Our study has limitations. First, this is a retrospective series without uniform and systematic collection of data for all patients. Second, as mentioned above, these patients are selected by virtue of their referral to a histiocytosis referral center, often for molecular evaluation and for consideration of targeted therapy. Additionally, these are patients referred for their neurologic disease, and therefore may furthermore not be representative of ECD patients generally. Last, we acknowledge that the clinical and radiologic response assessment methodology employed is subjective, albeit published in ECD studies, and does not conform to established criteria applied within neuro-oncology. We believe that these limitations may also present a strength of this study. Our cohort is enriched for neurologic ECD (and for having undergone dedicated neurologic evaluation), which provides opportunity to examine these aspects of ECD in greater detail and provide new ideas for the clinical evaluation of other ECD patients. Also, the selection of our patients for molecular evaluation has allowed us to demonstrate a variety of non-BRAFV600E mutations in CNS ECD and their responses to various therapies.

We have presented 30 patients with neurologic ECD, demonstrating a wide spectrum of disease manifestations and outcomes. Our data suggest underrecognized symptomatology as well as favorable responses to targeted therapies. Our data further suggest that neurodegenerative phenomena may be an important element of the clinical neurologic ECD constellation that will benefit from further study.

Funding

This work was supported by the Erdheim-Chester Disease Global Alliance (E.L.D.) and National Institutes of Health/National Cancer Institute Core Grant (P30 CA008748). This work was also supported by the Frame Fund and Joy Family West Foundation (E.L.D.).

Conflict of interest statement

Dr Ulaner discloses consultant fees or research support from Sanofi, Genentech, and Puma biotechnology, unrelated to the current work.

Dr Rampal discloses consultant fees from Incyte, Celgene, Agios, Jazz, BeyondSpring, Apexx, and Partner Therapeutics, and research funding from Stemline and Constellation, unrelated to the current work.

Dr Hyman has served as a consultant or in an advisory role for Chugai Pharma, Boehringer Ingelheim, AstraZeneca, Pfizer, Bayer, Genentech, and Fount. Dr Hyman has stock and/or other ownership interest in Fount and has received research funding from AstraZeneca, Puma Biotechnology, Loxo, and Bayer.

Dr Abdel-Wahab has served as a consultant for H3B Biomedicine, Foundation Medicine Inc, Merck, and Janssen and serves on the Scientific Advisory Board of Envisagenics Inc; O.A-W. has received prior research funding from H3B Biomedicine unrelated to the current manuscript.

Dr Dogan receives consulting fees from Celgene, Corvus Pharmaceuticals, Novartis, Roche, Seattle Genetics, and Takeda, and research funding from Roche, all outside the submitted work.

Dr Panageas reports stock ownership in Johnson & Johnson, Viking Therapeutics, Catalyst Biotech, and Pfizer, outside of the submitted work.

Dr Diamond reports nonfinancial support from Third Rock Ventures, unrelated to the current work.

A Bhatia, V Hatzoglou, G Ulaner, R Rampal, D Hyman, O Abdel-Wahab, B Durham, A Dogan, N Ozkaya, M Yabe, K Petrova-Drus, KS Panageas, AS Reiner, M Rosenblum, E Diamond analyzed and interpreted the data.

A Bhatia, V Hatzoglou, G Ulaner, R Rampal, D Hyman, O Abdel-Wahab, B Durham, A Dogan, N Ozkaya, M Yabe, K Petrova-Drus, K Panageas, A Reiner, M Rosenblum, E Diamond wrote the manuscript.

All authors approved the final manuscript.

Authorship statement

A Bhatia, V Hatzoglou, G Ulaner, O Abdel-Wahab, B Durham, A Dogan, N Ozkaya, M Yabe, K Petrova-Drus, M Rosenblum, E Diamond collected the data.