Presented in part at the 60th Meeting of the American Academy of Neurology, Chicago, Illinois, April 12-19, 2008.
Intracranial dural metastases (IDM) are found at autopsy in 9% of patients with advanced systemic cancer. However, to the authors' knowledge, IDM have not been studied systematically in the modern neuroimaging era. The objective of the current study was to evaluate the demographics, clinical presentation, imaging, treatment, and prognosis of patients with IDM.
The current study was a retrospective review of 122 patients with IDM diagnosed at Memorial Sloan-Kettering Cancer Center between 1999 and 2006. Patients with concurrent brain or leptomeningeal metastases were excluded.
Sixty-one percent of the patients were women; the median age at diagnosis was 59 years, the median Karnofsky performance scale (KPS) at diagnosis was 80, and the median time to IDM diagnosis from initial cancer diagnosis was 37 months. Breast (34%) and prostate (17%) cancers were the most frequent primary tumors associated with IDM. Fifty-six percent of patients had a single dural metastasis. On imaging, 70% had metastases of the overlying skull, 44% had dural tail metastases, 53% had vasogenic edema, and 34% had brain invasion. Direct extension from skull metastases was the most common mode of spread. Eighty-three percent of patients had active systemic disease at the time of IDM diagnosis. A lower KPS and lung cancer were associated with worse overall survival. Surgical resection and chemotherapy improved progression-free survival, but only resection was found to be associated with improved overall survival.
Intracranial metastases are the most common neurologic complication in cancer patients. They are a significant cause of morbidity and mortality and reduce the quality of life due to progressive neurologic impairment. A variety of intracranial sites can be involved by metastatic tumor. Brain parenchyma is the most common, but the leptomeningeal and pachymeningeal (dural) compartments can also be involved. Dural metastases include those to the epidural space, usually direct extension from skull metastases, and those to the subdural space, either by direct extension from an epidural metastasis or by hematogenous spread.1
Intracranial dural metastases (IDM) occur in up to 9% to 10% of all patients with systemic cancer based on autopsy reports from 1950 to 1976.2, 3 In 1 of these autopsy series, there were 105 cases (4%) in which dural metastases represented the only site of intracranial tumor involvement.2 Although these numbers are fairly substantial and comparable to the incidence of leptomeningeal metastases, IDM have not been studied systematically in the modern neuroimaging era; recent studies have been case reports, small patient series, or reviews of published cases.4, 5 Furthermore, many of these studies have included patients with IDM and concomitant brain parenchymal or leptomeningeal metastases, and this limits a clear interpretation regarding the effect of IDM on prognosis and treatment. Moreover, the incidence of intracranial metastases may be increasing as therapies for systemic cancer improve and patients survive longer.1, 6 Therefore, we sought to evaluate the demographics, clinical presentation, radiographic characteristics, treatment, and prognosis of patients with IDM without leptomeningeal or intraparenchymal brain involvement.
MATERIALS AND METHODS
We retrospectively reviewed the charts and neuroimaging studies of patients with IDM who were evaluated by the Neurology Service at Memorial Sloan-Kettering Cancer Center (MSKCC) from 1999 to 2006. Inclusion criteria were: 1) diagnosis of IDM based on radiographic appearance (IDM identification was confirmed by reviewing the brain magnetic resonance imaging [MRI] and computed tomography [CT] scans of all patients.); 2) histologic diagnosis of primary cancer reviewed at MSKCC; 3) negative cerebrospinal fluid (CSF) cytology if lumbar puncture was performed; and 4) if resection of the IDM was performed, histology was to be compatible with the systemic cancer diagnosis. We excluded patients with concomitant brain or leptomeningeal metastases, determined either radiographically or by positive CSF cytology. We screened 192 patients with possible dural metastases, and 122 patients fulfilled the inclusion criteria; 34 had brain metastases, 12 had leptomeningeal metastases, and 24 had other diagnoses. This study was approved by the MSKCC Institutional Review Board.
Progression-free survival (PFS) was calculated from the date of IDM diagnosis until any further progression of disease (systemic or intracranial), death, or last follow-up. Overall survival (OS) was calculated from the date of IDM diagnosis until death or last follow-up. Potential prognostic factors and treatment modalities were evaluated using Cox proportional hazards analyses. Multivariate analyses were performed to identify variables that were found to be independent predictors of outcome. Factors with a P value ≤.1 in the univariate analyses were entered as candidate variables, and multivariate analyses were thus performed in a stepwise fashion. Calculations were performed using Stata statistical software (version 8.0; StataCorp, College Station, Tex).
One hundred twenty-two patients (61% of whom were women) with a median age at diagnosis of IDM of 59 years and a median Karnofsky performance scale (KPS) of 80 (Table 1) were included. The median time to IDM diagnosis from the initial cancer diagnosis was 36.6 months; 11% of patients presented with IDM at the time of initial cancer diagnosis.
The most common primary sites of cancer were the breast, prostate, lung (small cell and nonsmall cell), and head and neck and hematologic malignancies such as lymphoma and leukemia. Other tumor types included renal cell cancer, neuroblastoma, thyroid cancer, thymic cancer, leiomyosarcoma, carcinoid, melanoma, squamous cell cancer of skin, and adenocarcinoma of unknown primary origin. Headache and cranial neuropathy were the most common presenting symptoms and signs, followed by visual changes, alterations in mental status, and seizures (Table 1). Eleven percent of the patients were asymptomatic and were diagnosed incidentally by neuroimaging. A total of 101 patients (83%) had active systemic disease at the time of IDM diagnosis; the most commonly involved sites were bone (76%), lymph nodes (55%), liver (33%), and lungs (30%).
Ninety-three percent of patients had gadolinium-enhanced brain MRI scans and 7% had head CT scans with intravenous contrast available for review. Fifty-six percent of patients had a single dural metastasis, and 25% had diffuse dural involvement (Fig. 1A). The most common sites were parietal (36%) and frontal (32%); 11% had infratentorial lesions. Lesions with a characteristic dural tail sign were found in 44% of patients (Fig. 1B). In 53% of the patients, there was compression of the underlying brain parenchyma with resultant vasogenic edema; 34% had actual brain invasion. Other radiologic findings included involvement of venous sinuses (Fig. 1C), which caused occlusion or compression of the sinus and structures passing through them such as cranial nerves in the cavernous sinus (Table 1). Subdural hematoma and effusions were noted in 3 patients.
Types of Dural Involvement
Dural involvement was a result of direct extension of skull metastases in 61% of patients (Fig. 1D) and of hematogenous spread in 33%. Six cases were possibly secondary to postsurgical seeding of the dura mater after complete resection of a brain metastasis; these patients did not have any parenchymal disease at the time of the IDM diagnosis. However, due to the location of dural metastasis at the surgical site of the prior resected brain parenchymal metastasis, we presume the surgical seeding was the most likely mode of spread. There was no significant difference with regard to age, sex, KPS, cancer histology, or clinical presentation between patients with dural involvement from skull extension versus those with other types of dural involvement, except that patients with IDM due to skull extension had seizures less frequently (5% vs 21%; P = .007) and cranial nerve involvement more often (37% vs 19%; P = .03).
The treatment was not standardized and included surgery, chemotherapy, radiotherapy (RT), a combination of treatment modalities, or supportive care only. Thirty (25%) patients underwent surgical resection, of whom 63% underwent a complete resection and 37% underwent a partial resection. Among the patients who underwent surgical resection, 15 patients received RT (11 focal and 4 whole-brain RT [WBRT]) and 10 received chemotherapy. Among all patients, 63 patients (52%) received some type of RT; of these, 47% underwent WBRT, 47% underwent focal RT, and 6% underwent stereotactic radiosurgery (SRS). Fifty-five (45%) patients received chemotherapy and 21 (17%) received supportive care only.
Outcome and Prognostic Factors
The median follow-up of the surviving patients was 60 months. At the time of last follow-up, 108 patients had died. The median PFS was 3.7 months (95% confidence interval [95% CI], 2.8-4.3 months), and the median OS was 9.5 months (95% CI, 6.6-12.5 months). One hundred patients had documented disease progression; the initial site of progression was systemic in 49 patients, intracranial in 30 patients, and both systemic and intracranial in 21 patients. Sixty-three patients had intracranial disease progression at some point in their disease course. The sites of intracranial progression included local in the dura in 54 (86%) patients, distant in the dura in 23 (37%) patients, intraparenchymal in 26 (41%) patients, and leptomeningeal in 4 (6%) patients.
On univariate analysis, older age, lower KPS, the presence of systemic disease, and lung cancer were found to be associated with worse PFS and OS, whereas resection and chemotherapy were associated with better PFS and OS (Table 2). On multivariate analysis, lower KPS and lung cancer were associated with worse PFS and OS, and resection was associated with better PFS and OS, whereas chemotherapy was found to improve PFS only (Table 3).
Table 2. Univariate Analysis of Progression-free Survival and Overall Survival
This large series of patients with IDM evaluated in the modern neuroimaging era provides insight into the common primary cancers associated with IDM, clinical and imaging characteristics, and treatment options. The median survival of patients with IDM was found to be 9.5 months, which compares favorably with reported survivals of 4 to 6 months for patients with brain metastases1 and 2 months for those with leptomeningeal metastases.7 In our series, breast cancer was the most common primary tumor associated with IDM, followed by prostate and lung cancer. Our numbers are comparable to the autopsy study by Takakura et al.,3 in which 31% of IDM patients had breast cancer, 11% had prostate cancer, and 7% had lung cancer. The higher incidence of breast cancer explains the predominance of women in the current study. Head and neck cancers were also frequent, because these tumors can invade the base of the skull and adjacent tissues such as the dura mater.8
Clinical presentation often suggested the location of tumor and extent of involvement. However, one of the most common presenting symptoms is headache, which may or may not be of localizing value. Headache may be indicative of traction on the dura, invasion of venous sinuses, or raised intracranial pressure. Compression of the cavernous sinus is associated with multiple cranial neuropathies. Focal weakness or sensory loss helped in localizing the tumor to the contralateral hemisphere and was due to either compression of the underlying brain by the dural tumor or brain invasion.5
Gadolinium-enhanced brain MRI is the study of choice for the diagnosis of IDM. The appearance of IDM varies and may appear simply as localized thickening or may be nodular, classically lenticular or biconvex, often resembling a meningioma. These may appear as either single or multiple metastases. Alternatively, there may be diffuse dural enhancement or diffuse involvement with nodular areas. The absence of a blood-brain barrier explains the intense and homogeneous contrast enhancement noted on MRI. Moreover, gadolinium-enhanced brain MRI is important in diagnosing concurrent leptomeningeal or brain parenchyma metastases, which has important therapeutic and prognostic implications. Whether an isolated meningeal lesion is a meningioma or a dural metastasis can be a clinical dilemma, particularly in women with breast cancer, who have an increased incidence of meningiomas.9-11 Several patients had a single lesion, with the characteristic dural tail sign making it radiographically indistinguishable from meningioma. In these patients, other features such as overlying skull metastases and bony erosion aided in the clinical diagnosis. In the absence of other characteristics, some patients underwent surgical resection with confirmation of tumor type by pathology; others were followed with repeat imaging before implementing therapy. Diffuse dural enhancement may also be challenging diagnostically. It may represent dural invasion by cancer or a reactive response of the dura mater to the skull metastases; infectious or inflammatory etiologies also need to be considered. A review of 198 cases published in the English and French medical literature between 1904 and 2003 found that 41% of patients with IDM had a subdural hematoma presentation,5 which was found in only 3 of the cases in the current study. This finding might be related to the bias of publishing unusual presentations. Subdural hematomas in this setting could be related to the neovascularization of dural metastases and possibly exacerbated by thrombocytopenia or coagulation disorders due to the underlying cancer or its treatment.
Direct extension from skull metastases was the most common mode of spread, which explains the high association with breast and prostate cancer, which metastasize commonly to bone.12 Dural involvement from the outward progression of cortical brain metastases is rare.5 However, surgical seeding of the dura after resection of a brain metastasis was observed in a few of our cases. In the current study, 76% of the patients had bone metastases, including skull lesions. Conversely, lung metastases (excluding primary lung lesions), which are reported in up to 67% of patients at the time of diagnosis of brain metastases, were found in only 30% of the patients in the current study.13 Consequently, hematogenous spread, which accounts for the majority of brain metastases, was reported to occur in only 33% of our IDM patients.
Age, performance status, active systemic disease, lung cancer, and number of metastatic lesions are known prognostic factors for patients with brain metastases.14-16 In our study of IDM, we found that lower KPS and a lung primary tumor were associated with worse survival; other factors may have appeared to have no effect due to the relatively small number of patients. Resection of IDM improved survival. Randomized trials have shown that surgical excision followed by RT is superior to RT alone in the treatment of single brain metastases.17, 18 Because the data from brain metastases cannot be extrapolated to IDM, the current study data may stimulate prospective studies to determine the role of surgical resection in patients with a single dural lesion. Neurosurgeons should be consulted to evaluate patients with IDM. Patients with a single resectable symptomatic IDM, controlled systemic cancer, and acceptable surgical risk should be considered for resection as first-line therapy.
IDM are outside the blood-brain barrier, which likely explains the beneficial effect of systemic chemotherapy on PFS. This is in contrast to brain or leptomeningeal metastases, in which the blood-brain barrier decreases central nervous system penetration and the effectiveness of systemic chemotherapy. Moreover, patients with IDM frequently have active systemic cancer that requires chemotherapy. Consequently, patients with IDM who are candidates for systemic chemotherapy should receive the best available regimen directed to the underlying cancer because blood-brain barrier penetration is usually not an issue.
RT given in the form of WBRT, focal RT, or SRS did not demonstrate a beneficial effect on survival in the current study. This could be related to selection bias because patients with worse performance status are often not candidates for surgical resection or aggressive systemic chemotherapy but still can undergo palliative RT. Furthermore, lesions in particular areas, such as the skull base, are not amenable to a surgical approach, but RT might still play a role.
The current study is retrospective and therefore has inherent limitations. In addition, the current study was conducted at a tertiary cancer center, and the patient population may not be representative. The treatment was not standardized, and therefore it was not possible to arrive at definitive treatment recommendations. However, to our knowledge, the current study is the largest nonautopsy series of IDM that was systematically evaluated from a single institution. Further studies are needed to clarify these issues.
We thank Judy Lampron for her expert editorial assistance and Dr. Lisa M. DeAngelis for her support and helpful suggestions.