|Home | About | Journals | Submit | Contact Us | Français|
The purpose of this study was to describe a cohort of patients with leptomeningeal melanomatosis (LM) and to determine prognostic factors for outcomes in these patients. The primary hypothesis was that more extensive burden of CNS metastasis at the time of diagnosis of LM (as evidenced by imaging of the CNS parenchyma and meninges and cerebrospinal fluid [CSF] cytology status [positive versus negative]) correlates with poorer outcomes. The records of all patients with LM treated at M. D. Anderson Cancer Center between 1944 and 2002 were reviewed. Information on clinical course and outcomes was gathered. Univariate and multivariate analyses were performed on 110 patients using Cox proportional hazards regression analysis to examine the effects of possible predictive factors on survival. The overall median survival from LM diagnosis was 10 weeks, with a 95% confidence interval (CI) of 8–14 weeks. Eighty-six (78.2%) patients had cutaneous primary lesions, and 23 (20.9%) had melanoma of unknown primary site. The primary hypothesis was not proven. Neither the presence of parenchymal CNS metastases, nor greater imaging evidence of LM, nor positive CSF cytology at diagnosis correlated with survival outcomes. Univariate analyses revealed possible predictors of longer survival, including the presence of supratentorial or spinal LM on imaging at diagnosis versus its absence and any treatment of LM, whereas elevated serum lactate dehydrogenase at the time of LM diagnosis predicted shorter survival. Multivariate analysis revealed that a history of a primary melanoma lesion originating on the trunk predicted shorter survival after LM diagnosis (hazard ratio [HR] = 2.0, 95% CI = 1.0–3.8, p = 0.035), and treatment with intrathecal chemotherapy predicted longer survival (HR = 0.5, 95% CI = 0.4–0.8, p = 0.0036). The positive result with respect to treatment is unreliable due to the inability to remove treatment selection bias from the analysis. This retrospective analysis confirmed the dismal prognosis associated with LM. The amount of CNS tumor burden at the time of diagnosis of LM did not inversely correlate with survival outcomes, contrary to our hypothesis.
Leptomeningeal metastasis from malignant melanoma, leptomeningeal melanomatosis (LM), portends a bleak prognosis. A compilation of 16 studies of patients with solid tumors and with neoplastic meningitis (NM), including patients with LM, revealed median survival times ranging from 8 to 16 weeks after diagnosis.1 There is evidence that the incidence of NM from various tumors is rising, likely due to improved systemic treatments and diagnostic testing.2 Breast and lung carcinoma, along with melanoma, have been identified as the most prevalent causes of NM from solid tumors.1 Small-cell lung cancer and melanoma have the highest rates of spread to the leptomeninges, at 11% and 23%, respectively.4,5
Brain metastases are found concurrently with NM in 28%–75% of solid-tumor NM patients.2,6 Similarly, with melanoma there appears to be a relationship between parenchymal CNS metastases and LM; 19% of patients with parenchymal CNS tumor also had LM in one study,7 whereas at autopsy only 5% of patients had LM without accompanying parenchymal CNS metastases.8
The literature contains no sizable study with prognostic indicators specific to LM. Studies of NM from grouped solid tumors (in which melanoma was a small subset in a few series)2,3,9–18 have rendered some useful prognostic indicators, along with equivocal and inconsistent results.
The purpose of this study was to assess the impact of CNS tumor burden (both brain metastases and LM as quantified by neuroimaging appearance, neuraxis level, and numbers of clinically affected neuroanatomic levels, as well as cerebrospinal fluid [CSF] analysis) on outcomes in LM. Secondary aims were to report the clinical features of LM and explore the data for other potential prognostic factors in LM patients.
Patients diagnosed with malignant melanoma with metastasis to the leptomeninges between 1944 and 2002 were identified from the database of the tumor registry at M. D. Anderson Cancer Center (MDACC, Houston, TX, USA). Inclusion criteria were that the patients must have had leptomeningeal metastasis from melanoma as evidenced by neuroimaging or CSF analyses, which were confirmed by MDACC faculty. The time of diagnosis was the date of the first piece of objective evidence of LM, whether that was obtained at MDACC or at another institution. The study was approved by the Institutional Review Board of The University of Texas MDACC.
Data collected from the subjects’ medical records included demographics, Breslow thickness, Clark level, histology of presenting lesion, location of presenting melanoma lesion, and time interval between diagnosis of primary melanoma and LM. Information on systemic tumor burden and serum lactate dehydrogenase (LDH) was recorded at the time of LM diagnosis. Clinical signs and symptoms of LM, along with CSF analysis and neuroimaging describing location of LM (supratentorial, infratentorial, spinal) and characteristics (nodular or diffuse enhancement of the leptomeninges) were obtained at the time of LM diagnosis. Neuroimaging from the time of LM diagnosis was reviewed when the radiology report was insufficient. The relationship of LM to parenchymal CNS metastasis was investigated, including parenchymal lesion location, number, and treatment. Each therapeutic intervention and response was documented for all stages of disease.
The potential prognostic factors evaluated were age, gender, ethnicity, year of diagnosis, site of primary lesion, time of primary diagnosis to LM, number and sites of systemic metastases, parenchymal tumor burden, LM tumor burden measured by neuroimaging and symptoms, and treatment. Death or last follow-up was the determined end point. Because this study was retrospective, selection bias was a presumed confounding factor; therefore evaluation of therapy for LM, although reported, should be considered in this light.
Univariate and multivariate statistical analyses were performed using Cox proportional hazards regression analysis to identify prognostic factors and the impact of treatment on survival after diagnosis of LM. Date of diagnosis of LM was the date on which a test (either imaging or CSF) identified the disease, and survival was measured from that date to the date of patient death or last follow-up. A p-value < 0.05 was considered significant. Kaplan-Meier curves were constructed to estimate survival percentages. The impact of all therapies (radio-therapy, systemic, and intrathecal chemotherapy) on survival were evaluated individually using multivariate techniques.
Of the 149 records identified by the database, 18 were not included due to insufficient diagnostic evidence of LM, and 2 were not available for review. Eight patients were diagnosed with LM at autopsy, and 4 were diagnosed within 1 week of death; these 12 were excluded from analysis. Seven patients presented with findings suggestive of primary meningeal melanoma and were excluded from analysis. Analysis was conducted in the remaining 110 patients.
Of the 110 patients, 106 (96.4%) had evidence of melanoma metastases other than LM; 4 patients (3.6%) presented with LM alone. Initial LM diagnosis was confirmed by CSF cytology alone in 28% (31 of 110) and by imaging alone in 33% (36 of 110). Both CSF cytology and imaging contributed to the diagnosis of LM in 38% (42 of 110). One diagnosis was established by leptomeningeal biopsy.
Diagnostic imaging methods changed over time. Out of 15 patients diagnosed before 1990, nine underwent CT brain imaging (useful in diagnosis in five patients), five patients underwent myelography (useful in four patients), and one brain and two spine MRIs were performed (all images useful in diagnosis). Two nuclear medicine scintigraphy scans, one arteriogram, and one ventriculogram were performed, and all were nondiagnostic. Beginning in 1990, of 95 patients, 5 (5%) had no imaging, 4 of the remaining 90 (4%) had head CTs only (2 studies diagnostic, 2 nondiagnostic), and 86 patients (95%) underwent MRI imaging of the brain and/or the spine. Over time, MRI imaging became nearly the exclusive imaging tool used in the diagnostic evaluation of patients suspected of having LM.
Fifty-four of 110 (49%) patients were diagnosed with parenchymal CNS tumor before the diagnosis of LM; 42 of 110 (38%) patients had no prior history of CNS metastases. Nine patients (8%) were diagnosed with parenchymal CNS metastases simultaneously with the LM diagnosis, and five patients (5%) went on to develop parenchymal CNS metastases after they were diagnosed with LM. Of the 54 patients with a prior CNS metastasis diagnosis, 12 had solitary brain metastases and 42 had multiple lesions. All patients received treatment for intraparenchymal tumor; 21 of 54 (39%) received radio-therapy to the brain and/or spine, and 38 of 54 (70%) had at least 1 craniotomy before LM diagnosis.
The majority (86 of 110; 78.2%) presented with a cutaneous primary lesion and 23 (20.9%) with melanoma of unknown primary site, 3 of which (2.7%) presented with leptomeningeal metastases at the time of the establishment of the diagnosis of melanoma. One patient (0.9%) had mucosal melanoma. It has been suggested that the 5-year survival rate for melanoma of unknown primary site does not differ from control groups of cutaneous melanomas;19 thus those with unknown primary sites were included in analysis. Details on patient demographics and characteristics of the primary melanoma are listed in Table 1. Over the same study interval, 23,169 melanoma patients were treated at MDACC, and 20,263 (87%) had cutaneous melanoma. Melanoma of unknown primary site was documented in 1,371 patients (6%). The remaining 1,592 (6.5%) patients had primary melanoma of other origin. For the 110 cases included in analysis, median interval from diagnosis of melanoma to LM was 3 years (Table 2).
LM is generally considered to be a late event in melanoma patients, and the extent of systemic disease may impact survival. In our patients, 81% had at least 2 other sites of metastatic melanoma besides their meningeal disease. The most commonly affected sites were the visceral organs (61% of patients), followed by the lymph nodes (50%) and the nonmeningeal nervous system (48%). Details on the extent of non-LM metastases are presented in Table 3. Serum LDH is sometimes used as a surrogate marker for the extent of systemic disease in melanoma. Serum LDH level was available for 67 of our patients; 32 of 67 (48%) had an elevated LDH level (>600 IU/liter).
Clinical features of LM at diagnosis are reported in Table 4. The most common presenting symptoms of LM included headache (46%), nausea or vomiting (37%), back or neck pain (24%), and weakness (22%). The most common presenting signs in newly diagnosed LM patients were sensory loss (23%), cranial nerve III, IV, or VI abnormality (18%), lower motor neuron weakness (18%), and mental status changes (17%).
Treatments that patients received are presented in Table 5. Beginning in 1990, a higher percentage of LM patients received no treatment for their LM than before 1990, including lower percentages of patients receiving radiotherapy and intrathecal chemotherapy. Before 1990, nitrosoureas were the most commonly prescribed systemic chemotherapy agents for LM, whereas beginning in 1990, procarbazine was used in 53% (20 of 38) of the patents who received systemic chemotherapy. After temozolomide was approved by the U.S. FDA for recurrent glioma in 1999, 8 of 15 (53%) chemotherapy-treated LM patients received this treatment. Intrathecal chemotherapy was used in a higher percentage of patients before 1990, the most commonly used agent being methotrexate. Beginning in 1990, intrathecal interleukin-2 was used in 76% (31 of 41) of patients treated with intrathecal agents.
The overall median survival from LM diagnosis was 10 weeks (95% confidence interval [CI] = 8–14 weeks; Fig. 1). One-year survival was 7% (95% CI = 4%–14%), and at 2 years it was 3% (95% CI =1%–8%). One patient is alive 11.2 years after LM diagnosis and continues to receive regularly scheduled intrathecal interleukin-2 for LM. All other patients included in this study have died.
There were no differences in survival in patients who had other CNS metastases versus those without. The 54 patients diagnosed with parenchymal CNS tumor before the diagnosis of LM had a median survival of 9 weeks, and the 42 patients without a prior history of CNS metastases had a median survival of 8 weeks after the diagnosis of LM.
Signs and symptoms of LM were grouped by anatomic locations (cerebrum, cranial nerve, and spine) and provided no prognostic information regarding survival (patients with a history of brain metastases were excluded from this analysis because their signs or symptoms could not definitively be attributed to LM). Performance status at diagnosis of LM was available from only 23 patients and therefore was not included in the analysis.
The presence or absence of visible LM on neuroimaging at LM diagnosis had no prognostic significance (p = 0.29). LM tumor burden measured by the number of tumor deposition sites (defined by supratentorial, infratentorial, cranial nerve, and spinal) along the neuraxis, as identified by neuroimaging, was associated with a nonsignificant (p = 0.11) increase in survival with increasing number of sites of involvement. Similarly, patients with malignant cells in the CSF had no significant difference in their survival times (median 12 weeks) when compared to patients without malignant cells (median 10 weeks). Further, combinations of positive or negative CSF results, when combined with positive or negative imaging tests, revealed no differences in survival times between groups.
Numbers of nonmeningeal metastases and their sites were not included in the analysis of potential prognostic factors due to the nonuniform pattern of patient testing and the likelihood of missing data. Elevated serum LDH at time of LM diagnosis, a surrogate marker of systemic disease burden, correlated with a poorer survival after LM diagnosis (hazard ratio [HR] = 1.8, 95% CI = 1.1–3.0, p = 0.019).
Univariate analysis revealed that all three modalities of treatment, radiotherapy (HR = 0.5, 95% CI = 0.4, 0.8, p = 0.0015), systemic chemotherapy (HR = 0.6, 95% CI = 0.4, 0.9, p = 0.028), and intrathecal chemotherapy (HR = 0.5, 95% CI = 0.3, 0.7, p = 0.0001), appeared to positively impact survival.
Multivariate analysis of potential prognostic factors, listed in Table 6, was performed using Cox proportional hazards regression modeling. Only factors for which there were adequate numbers of data points were included in the analysis. The only significant predictor of survival was that of cutaneous primary lesion of the trunk, with a hazard ratio of 2.0 (p = 0.035), associated with a poorer survival after LM diagnosis. A majority of patients received some form of therapy directed toward their LM (82 of 110, 75%). After multivariate analysis, intrathecal chemotherapy was significantly associated with longer survival, and radiotherapy and systemic chemotherapy were not protective, but they were not associated with detrimental effects (Table 7).
Leptomeningeal melanomatosis is an incapacitating complication of melanoma. With previously reported poor response to conventional intrathecal chemotherapy and radiotherapy19–23 and variable responses and neurotoxicity associated with intrathecal immunotherapy,24–29 reliable prognostic factors are needed to guide therapeutic decision making.
One surprising finding in this study was the low incidence of LM (0.56%). The reasons for this low percentage are not clear, although they may be secondary to missed diagnoses from limited tests available in the early years of this study, continued care at other institutions after consultation at MDACC, lack of inclusion of the diagnosis of LM in our database, or perceived futility of treatment, restricting testing and establishment of the diagnosis.
Up to 60% of patients with metastatic melanoma have clinically apparent brain metastases.30 In a study of patients with NM from solid tumors, prior history of intraparenchymal CNS tumor was associated with longer progression-free survival.9 CNS metastases were a common site of metastases in the presence of LM in this series (48%), yet this factor did not correlate with outcome, nor did the number of parenchymal lesions. Notably, all patients with CNS parenchymal disease before LM underwent surgery, radiation, or chemotherapy, plausibly suppressing dissemination of melanoma into the CSF. Additionally, many patients underwent periodic neuroimaging, potentially facilitating earlier detection of LM, and artificially extending survival time when compared to patients in whom LM was the first CNS site of metastasis. These confounding factors may have countered the negative effects of parenchymal metastases. Parenchymal metastases were not important in prognostication, as hypothesized.
The absence of visible bulky CNS tumor (parenchymal or leptomeningeal) on neuroimaging has independently predicted improved survival in studies of mixed solid tumors.9,31 This finding was not confirmed in the present study. In univariate analysis, the presence of visible LM on neuroimaging did not adversely affect survival as hypothesized. A number of variables may have influenced this finding, including differences in imaging technique and imaging completeness between patients (complete neuraxis imaging not obtained in all patients). Further, the study technique and sensitivity for disease detection may have varied between patients and over time.
Limitations inherent to the methods of quantifying LM may have confounded the results. One report in the literature describes MRI findings obtained within 12 h of death from LM, followed by autopsy. The autopsy demonstrated that subtle MRI findings did not reflect the black tumor coating the surface of the entire neuraxis.32 This report demonstrated potential discrepancies between MRI characteristics and actual leptomeningeal tumor burden, such that neuroimaging may not be an accurate measure of LM tumor. In addition, the imaging evidence of disease may not correlate with the biological aggressiveness of the disease.
The positive impact of supratentorial LM on survival (p = 0.017, univariate analysis only) may be the result of earlier diagnosis related to cranial symptoms that prompted earlier evaluation. Additionally, prior treatment directed at the cranial structures for brain metastases may have lengthened survival in patients with supratentorial LM.
This report provides percentages of signs and symptoms of patients with LM at presentation, which are similar in frequency to those of reports of NM from multiple histologies.6 Prior studies evaluating the significance of signs and symptoms as they relate to prognosis have yielded mixed results.2,15,17 In this study, we found no prognostic significance of the neuraxis level of signs and symptoms.
The prognosis of stage IV melanoma from time of distant metastases is 6–7.5 months, with a 5-year survival of <10%.33,34 As demonstrated in this and previous series,2,6 LM commonly occurs late in the course of disease (96.4% in this series), often coexisting with widespread dissemination.
Prognostication based on systemic disease has been investigated in brain metastases from melanoma. Sampson et al. reported that the most significant predictor of survival in patients with brain metastases from melanoma is the number of organs with metastatic lesions.35 Although this relationship has not been proven for LM, the present series suggests that it may be applicable. Notably, four cases of LM (3.6%) presented without identifiable coexistent systemic metastases and demonstrated a remarkably longer median survival of 58 weeks (range 23–140 weeks). Further analysis was not carried out in the current study because of nonuniform patient testing and potential missing data.
Elevated serum LDH is one of the most predictive independent factors in stage IV melanoma,36,37 but only a slight difference in survival time was demonstrated in this series, and only in univariate analysis.
Reporting the success or failure of treatment modalities for LM was not the primary aim of this study. Numerous factors were not controlled in this retrospective review, including multiple concurrent treatment modalities and discontinuations and interruptions of treatment. Although survival was universally poor, intrathecal therapy did have a positive effect on outcome in multivariate analysis; however, patient selection bias cannot be excluded from the perceived benefit, casting doubt on this result. Our group previously reported on modest positive effects of intrathecal IL-2 on outcomes in this patient population,38 and the longest-lived survivor reported here was treated with intrathecal IL-2.
Head, neck, and trunk cutaneous melanomas are associated with worse prognosis than cutaneous melanomas of the extremities.39 Melanomas of the trunk have also been associated with the development of brain metastases.36 Accordingly, worse prognosis for patients with lesions of the trunk in this series was not unexpected. Recent evidence suggests there are different mutations in tumors that originate in sun-exposed versus non-sun-exposed areas,40 which could be a hidden contributor to outcomes in the current study. Even in those few prognostic indicators identified, differences in survival were measured in weeks rather than in months.
Outcomes for patients with LM are extremely poor. Signs and symptoms of LM are similar to those in patients with NM from other histologies. The extent of CNS disease does not appear to influence survival. To improve outcomes for these patients, further advances are required, targeting both the systemic and CNS disease.
We appreciate the work of Sarah Taylor, manager of Medical Informatics, for detailed information on the patient populations. The patient population was identified through a search of the Tumor Registry database maintained by the Department of Medical Informatics. The authors are grateful for the suggestions and manuscript review of Dr. Nicholas E Papadopoulos from the Department of Melanoma at M.D. Anderson Cancer Center.