Neoplastic meningitis in patients with adenocarcinoma of the gastrointestinal tract


  • Pierre Giglio M.D.,

    1. Department of Neuro-Oncology, The University of Texas M. D. Anderson Cancer Center Brain Tumor Center, Houston, Texas
    Search for more papers by this author
  • Jeffrey S. Weinberg M.D.,

    1. Department of Neurosurgery, The University of Texas M. D. Anderson Cancer Center Brain Tumor Center, Houston, Texas
    Search for more papers by this author
  • Arthur D. Forman M.D.,

    1. Department of Neuro-Oncology, The University of Texas M. D. Anderson Cancer Center Brain Tumor Center, Houston, Texas
    Search for more papers by this author
  • Robert Wolff M.D.,

    1. Department of Gastrointestinal Medical Oncology, The University of Texas M. D. Anderson Cancer Center Brain Tumor Center, Houston, Texas
    Search for more papers by this author
  • Morris D. Groves M.D.

    Corresponding author
    1. Department of Neuro-Oncology, The University of Texas M. D. Anderson Cancer Center Brain Tumor Center, Houston, Texas
    • Department of Neuro-Oncology, The University of Texas M. D. Anderson Cancer Center, 1400 Holcombe Boulevard, Unit 431, Houston, TX 77030
    Search for more papers by this author
    • Fax: (713) 794-4999



Neoplastic meningitis (NM) occurs in 5–10% of patients with malignant disease. Little is known about the outcomes of patients with gastrointestinal (GI) malignancies who develop NM. For this report, the authors characterized the clinical course and attempted to identify prognostic factors in patients with NM due to primary malignancies of the GI tract.


In this retrospective study, 21 patients with GI primary tumors and NM were identified: Their medical records and imaging studies were reviewed.


The patient population was composed of patients with gastric adenocarcinoma (n = 8 patients), esophageal adenocarcinoma (n = 7 patients), colon and/or rectal adenocarcinoma (n = 5 patients), and pancreatic adenocarcinoma (n = 1 patient). The median overall survival after the initial diagnosis of adenocarcinoma was 55 weeks (range, 8–884 wks), and the median survival after the diagnosis of NM was 7 weeks (range, 0–64 wks). Four patients died during palliative radiotherapy. No factors identified had an impact on outcome, including symptoms, physical findings at diagnosis, imaging characteristics, or cerebrospinal fluid findings. Univariate analysis showed a trend toward better outcomes for patients who received any kind of treatment directed toward the NM.


Patients with NM from GI tract adenocarcinomas universally had poor outcomes. Until NM can be diagnosed earlier and/or until more effective therapies are identified, comfort care alone may be a reasonable alternative for some of these unfortunate patients. Cancer 2005. © 2005 American Cancer Society.

Neoplastic meningitis (NM) develops in 5–10% of patients with solid tumors1 and often occurs in patients who have tumors that result in distant systemic metastases. The development of NM portends a significant worsening of prognosis, with survival ranging from 4 weeks to 16 weeks after its diagnosis.1 Despite attempts to improve survival with intrathecal and systemic chemotherapy as well as radiation therapy, the outlook for patients with this condition remains grim. NM is seen most often in patients with breast and lung carcinomas and melanoma. Because NM is identified less commonly in patients with gastrointestinal (GI) and other solid tumors, information about NM in such patients is sparse. In an attempt to add to this paucity of knowledge, we investigated patient and primary disease characteristics, treatment strategies, and survival data for a series of patients with NM arising from primary tumor sites within the GI tract. The objectives of this study were to provide historic data that could be compared with newer interventions and to elucidate potential prognostic factors that can be used to improve therapeutic decision-making.


An institutional protocol (RCR03-0331) for retrospective chart analysis was prepared and approved by The University of Texas M. D. Anderson Cancer Center (MDACC) Institutional Review Board. Waiver of written informed consent was granted, as requested, with protocol approval. In total, 879 patients who had been seen at MDACC between 1944 and 2002 and diagnosed with metastases to the spinal and/or cranial meninges were identified in the data base of the Tumor Registry of the Medical Informatics Office at MDACC. Of these patients, 31 had a primary diagnosis of adenocarcinoma of the GI tract. Information on these patients was extracted from electronic and paper medical records, and imaging studies were reviewed. Patient demographic data, age at diagnosis of the primary tumor, and date of death were obtained. Information extracted about the primary tumor included the extent of disease and treatments received at the time of diagnosis of the primary tumor and when NM was diagnosed. Data regarding NM were collected and included signs and symptoms at presentation, mode of diagnosis of NM, extent of disease, treatment, and treatment complications. We also looked for the presence or absence of metastases in structures near the meninges, such as vertebral bodies. For patients who were diagnosed with parenchymal brain metastases, their number, location, and treatment also were noted.

Overall survival was calculated from the time of diagnosis of the primary tumor until death or last followup. Survival from the time of diagnosis of NM until death also was determined. Survival was estimated using the Kaplan–Meier method. Possible prognostic variables were assessed, including age, gender, disease stage at diagnosis of the primary tumor, time from primary tumor diagnosis to the diagnosis of NM, presence of parenchymal brain metastases with NM, and NM treatment (intrathecal chemotherapy and/or radiation therapy). Data analysis was performed using Statistica software (version 6.0; Statsoft Inc., Tulsa, OK). All statistical comparisons were made with 2-tailed tests, and P values < 0.05 were considered as significant.


Patient Characteristics

Thirty-one patients were identified in the NM cohort who had metastases to the meninges of the brain and/or spinal cord, and 21 patients were identified for statistical analysis. Ten of those patients were excluded from further analysis. In 8 of those 10 patients, the diagnosis of NM was made on the basis of clinical evidence only, without imaging studies or cerebrospinal fluid (CSF) corroboration. The remaining 2 patients were diagnosed with dural metastases rather than NM. There were 14 male patients and 7 female patients. The median age at the time the primary tumor was diagnosed was 51 years (range, 22–71 yrs).

Pathologic Details/Treatment of the Primary Tumor

All patients had adenocarcinoma, including eight patients with gastric adenocarcinoma, seven patients with esophageal adenocarcinoma, five patients with colon and/or rectal adenocarcinoma, and one patient with pancreatic adenocarcinoma. The frequency of occurrence of NM for each primary tumor site was as follows: gastric carcinoma, 8 of 5618 patients (0.14%) who were seen during the study period and 8 of 4126 patients (0.19%) who developed at least 1 metastasis during the study period; esophageal carcinoma, 7 of 4361 patients (0.16%) and 7 of 2737 patients (0.25%), respectively; colon or rectal carcinoma, 5 of 25,755 patients (0.019%) and 5 of 18,006 patients (0.027%), respectively; and pancreatic carcinoma, 1 of 6052 patients (0.016%) and 1 of 4315 patients (0.023%), respectively. Eleven patients had tumors described as signet ring cell tumors, 4 patients had moderately differentiated tumors, and 9 patients had poorly differentiated tumors. The details for each primary tumor, including location, stage at diagnosis, treatment at diagnosis, and recurrence, are presented in Table 1.

Table 1. Primary Gastrointestinal Carcinoma and Treatment Details in 22 Patients who Developed Neoplastic Meningitis
Patient no.Yr diagnosedAge (yrs)GenderTumor locationDistant metastasis at diagnosisTreatment of primary tumor at diagnosisTreatment of tumor at recurrence
  • CRT: chemoradiation; 5-FU: 5-fluorouracil; post-op: postoperative; Pre-op: preoperative; RT: radiation therapy; FAM: combined 5-FU, doxorubicin, and mitomycin-C; CI: continuous infusion; CA-in-situ: carcinoma in situ; CCNU: lomustine; AMSA: amsacrine (acridinyl anisidide); PALA; N-(phosphonacetyl)-L-aspartic acid.

  • a

    Denotes treatment according to institutional protocol.

  • b

    Neoplastic meningitis was present at the time the primary tumor was diagnosed.

1199571MaleEsophagusNoPreoperative CRT (5-FU); 5-FU post-opNone
2199966MaleEsophagusNoSurgical resectionNone
3200065MaleEsophagusNoPre-op CRT (5-FU + cisplatin)None
4200166MaleEsophagusNoPre-op CRT (CPT-11 + cisplatin) followed by 5-FU, taxol and RTaNone
5200261MaleEsophagusNo5-FU, carboplatin, + taxolNone
6200239MaleEsophagusYesbTaxol, carboplatin, + capecitabineNone
7200251MaleEsophagusNoPreoperative CRT (5-FU + cisplatin)None
8198052MaleStomachYesFAM CT + RT to gastric bedNone
9198766FemaleStomachYes5-FU, doxorubicinMitomycin + 5-FU
10200128MaleStomachNoPreoperative CRT (5-FU, cisplatin, folinic acid)None
11199849FemaleStomachYes5-FU, cisplatinum, taxolNone
12199947MaleStomachYesCisplatin, 5-FU, docetaxelaNone
13200035FemaleStomachYesCisplatin, epirubicin, 5-FUPaclitaxel and Cl of 5-FU
14200151FemaleStomachYesCisplatin, epirubicin, 5-FUNone
15200149FemaleStomachYesEpirubicin, cisplatin and 5-FU 
16197830MaleRectumNoSurgery for CA-in-situSurgery for 3 recurrences
17197960MaleRectumYesAbdominoperineal resection with ileostomy; vincristine, CCNU, 5-FU startedRT for recurrence followed by AMSA treatmenta; treated with PALAa and mitomycin-C for third and fourth recurrences
18199834FemaleColorectalNoPreoperative CRT (5-FU)None
19198638MaleColonNoSurgery + 5-FUSurgery + RT + 5-FU
21199253FemalePancreasYesb5-FU, leucovorin, cisplatinDoxorubicin + cytoxan

Details at Time of Clinical Presentation

The median patient age at the time of diagnosis with NM was 52 years (range, 27–72 yrs). The median time from diagnosis of the primary tumor to the diagnosis of NM was 39 weeks (range, 0–870 wks). Two patients, one with esophageal carcinoma and one with pancreatic carcinoma, were diagnosed with NM at the same time that their primary tumors were diagnosed. Their clinical presentation was typical of patients with NM. Headache was the most common symptom, and visual complaints were the next most common symptom. The overall symptom frequency is presented in Table 2. Physical examination showed papilledema in 5 patients (23.8%). Multiple cranial nerves were involved in 4 patients (19%).

Table 2. Type and Frequency of Symptoms in 22 Patients with Neoplastic Meningitis from Gastrointestinal Adenocarcinoma
SymptomNo. of patients (%)
Headache11 (52.0)
Visual symptoms8 (38.0)
 Visual blurring1
 Field deficits2
Vertigo/dizziness7 (33.0)
Nausea/emesis7 (33.0)
Focal symptoms6 (28.5)
Seizures5 (23.8)
Spinal symptoms4 (19)
Gait problems3 (14.3)
Neck pain3 (14.3)
Hearing loss1 (4.8)
Confusion1 (4.8)

Extent of Primary Tumor at the Time of Diagnosis of NM

Among the patients presented here, six patients had metastases to the brain parenchyma, skull base, or bony spine. Of 15 patients without known central nervous system (CNS) or para-CNS lesions, 8 patients had evidence of lymph node tumor spread, and 4 other patients had known metastatic disease to organs other than the lymph nodes. Three patients developed NM without evidence of metastatic disease to other organs. These included one patient each with gastric, esophageal, and colorectal carcinoma.

Establishing the diagnosis of NM.

The diagnosis was established from CSF cytology results after imaging studies in 12 patients. Nine patients did not undergo a lumbar puncture. Eight patients had characteristic brain and/or spine imaging findings, which alone established the diagnosis. In those patients who had both neuroimaging and CSF analysis at the time of diagnosis (n = 12 patients), all 12 patients had cytologic evidence of malignancy in the CSF, and 8 of 12 patients (67%) had neuroimaging tests that were suggestive of NM. In one patient, the diagnosis was made at autopsy. Details on CSF results are presented in Table 3, with only partial information available for some of the patients (n = 5 patients).

Table 3. Results of Cerebrospinal Fluid Testing
ParameterNo. of patients with abnormal findings (%)Median value (range)
  1. NA: not applicable.

Opening pressure (cm H2O)7/7 (100)32 (22–56)
Cytology (positive/negative)9/10 (90)NA
Glucose (mg/dL)4/8 (50)42 (26–78)
Protein (mg/dL)6/9 (67)102 (23–520)
Cell count7/7 (100)32 (11–131)

The imaging modalities used for diagnosis included computed tomography (CT) and magnetic resonance imaging (MRI) scans of the brain and/or spinal cord. Six patients had brain CT scans only without MRI. Of the six patients who had CT scans only, three patients had scans with contrast administered, and three patients had noncontrast-enhanced scans only. All patients who underwent MRI scans had contrast material administered with the scans. The criteria we considered diagnostic or suggestive of NM included diffuse or nodular enhancement of the surface of the brain or spinal cord or spinal or cranial nerves. Hydrocephalus was not used as diagnostic of NM. Only nine patients in this study underwent imaging of both the brain and the spine. Abnormalities suggestive of NM were seen in both areas in 5 of 9 patients (56%).

In 4 of 6 patients who had CT scans only (67%), the scans were normal, and a diagnosis was established from CSF cytology results. The CT scan from one patient revealed a questionable area of attenuation in the posterior fossa, and the NM diagnosis was confirmed at autopsy. Contrast CT scanning showed marked enhancement of the tentorium cerebelli and blastic changes in the region of the clivus in one patient, with the diagnosis subsequently confirmed by CSF cytology. NM was identified in 3 patients from CT scans of the brain followed by MRI imaging, and the CT scans were abnormal in 2 of those 3 patients (67%). After 1986, MRI scans were obtained for all patients. Depending on their clinical presentation, some patients were evaluated with both a brain and spine MRI with contrast. In all 16 patients who underwent MRI scans, abnormalities suggestive of NM were identified. Overall, disease was seen most commonly in the posterior fossa; it was noted in 10 patients. MRI findings are summarized in Table 4. CSF radionuclide flow studies were performed on only three patients, and no CSF blocks were identified.

Table 4. Magnetic Resonance Image Findings
Patient no.Location of imaging abnormalityCharacter of imaging abnormality
  1. ST: supratentorial; IT: infratentorial; CN: cranial nerve; DE: diffuse enhancement.

1  SpineNodular
2ST, ITCN Nodular and DE
3ITCNSpineNodular and DE
4  SpineNodular
10ST, IT  DE
11ST, IT SpineDE
13IT SpineDE
14ST, IT SpineDE
15ST  DE
18 CN DE
20IT SpineNodular and DE

History of brain metastases and disease sites contiguous to the meninges.

Six patients (28.5%) had either brain metastases or bony disease in the calvarium or spine at the time of NM diagnosis. Four patients had brain metastases. In two patients, skull base metastases were noted on imaging, and one of those patients also had vertebral metastases.

Treatment of NM.

There was no established treatment paradigm in place at MDACC over the period included in this study for patients with NM due to GI metastases. Accordingly, study patients were treated by different physicians using various approaches: Nine patients received radiation alone, four patients received intrathecal chemotherapy, two patients received both radiation and intrathecal chemotherapy, one patient received systemic chemotherapy, one patient received both radiation and systemic chemotherapy, and 4 patients received no treatment. Further details are presented in Table 5.

Table 5. Treatment of Neoplastic Metastases
Patient no.Treatment of NMRadiation detailsSurvival after diagnosis of NM (weeks)
  • NM: neoplastic meningitis; WBRT: whole brain radiation therapy; Gy: gray; IT; intrathecal; Fx: fractions; T: thoracic vertebrae; L: lumbar vertebrae; Ara-C: cytarabine; C: cervical vertebrae; S: sacral vertebrae; IV: intravenous; BCNU: carmustine; 5-FU: 5-fluorouracil.

  • a

    Treatment was interrupted due to patient death.

  • b

    This patient was reported previously (see Mallory et al., 195112).

2RadiationWBRT 22 Gy/11 Fxa8
3RadiationWBRT 9.26 Gy; 2.5 Fx/spine 15 Gy/5 Fxa1
4RadiationT10–L5 30 Gy/10 Fx15
5IT chemotherapy (topotecan)None6
6IT chemotherapy (methotrexate); then topotecan; then Ara-CNone20
7Systemic chemotherapy (capecitabine)None28
8RadiationWBRT 21 Gy/7 Fxa2
9RadiationC5 and LS spine (centered at L5–S1); 30 Gy/10 Fx to both fields14
10RadiationWBRT 30 Gy/10 Fx38
11RadiationWBRT 20 Gy/5 Fx7
13IT chemotherapy (mafosfamide)None2
14IT chemotherapy (methotrexate)None9
16bRadiation/IT chemotherapy (methotrexate)/IV BCNUWBRT 30 Gy/10 Fx; later, spine C5–T2, 30 Gy14
18Radiation/IT chemotherapy (methotrexate)WBRT 30 Gy/10 Fx8
19Radiation and systemic chemotherapy (5-FU)WBRT completed 5 Fx of a planned 30 Gy/10 Fxa1
20RadiationC2–C7 30 Gy/12 Fx7
21RadiationWBRT 30 Gy/10 Fx64

Additional studies and complications.

All six patients who received intrathecal chemotherapy underwent ventricular access catheter placement. Three patients had elevated CSF pressure that required intervention: One patient required repeated reservoir taps, and the two patients required changing their reservoirs to a ventriculoperitoneal shunt with an on-off valve. One patient developed meningitis caused by Staphylococcus epidermidis while receiving intrathecal methotrexate. One patient developed diarrhea due to oral capecitabine therapy.


The overall median survival for this group of patients after the diagnosis of GI carcinoma was 55 weeks (range, 8–884 wks). The median survival after the diagnosis of NM was 7 weeks (range, 0–64 wks). Survival after the diagnosis of NM for each tumor type was as follows: Patients with primary gastric tumors (n = 8 patients) had a median survival of 4.5 weeks (range, 2–38 wks); patients with primary esophageal tumors (n = 7 patients) had a median survival of 8 weeks (range, 0–28 wks); patients with primary colorectal tumors (n = 5 patients) had a median survival of 5 weeks (range, 0–14 wks); and patients with primary pancreatic tumors (n = 1 patient) had a median survival of 64 weeks.

Prognostic Factors

Univariate and multivariate statistical analyses were performed. Using chi-square tests, we found no differences in survival outcome proportions based on primary disease site, degree of tumor differentiation, age group (ages < 40 yrs, 41–55 yrs, and > 55 yrs), year of diagnosis (before 1981, from 1981 to 1999, and after 1999), number or location of areas in the CNS affected on imaging studies, character of imaging abnormality (diffusely enhancing or nodular), or treatment delivered. Using Pearson product-moment correlations, we found no association between survival after diagnosis of NM and age at diagnosis, Karnofsky performance status at diagnosis (data were available on only nine patients), year of diagnosis, the time from the original diagnosis to the development of NM, or the CSF protein or glucose level at the time of the original lumbar puncture. Two-tailed t tests revealed no differences in outcome based on the number or nature of the clinical signs or symptoms. The log-rank test revealed a trend (P = 0.07) and the Wilcoxon Mann–Whitney test revealed a significant difference (P = 0.01) in survival between patients who received treatment (regardless of the treatment; median survival, 7 wks) and patients who did not receive treatment (median survival, 1 week). NM was diagnosed at autopsy in one of the patients who received no treatment. Cox proportional hazards regression modeling identified no statistically significant interactions that had an impact on survival.


Despite the development of various antineoplastic therapies and the gradual improvement in patient survival, the need for an effective treatment for NM remains. Anatomically, the nervous system provides relative sanctuary for malignant cells from some antitumor therapies. Consequently, NM continues to develop in a significant number of patients with malignant disease and worsens their prognosis substantially. Retrospective data have shown an overall survival of only 4–16 weeks in patients with malignancies who develop NM.1 In an attempt to provide focused data on patients with NM from GI malignancies, in the current study, we looked at their clinical characteristics and survival. Our ultimate objective was to derive prognostic factors from our data that could be used to direct therapy and to generate survival curves for comparisons with future interventions.

Although we found that headache and visual symptoms were the most common presenting symptoms, no single symptom was useful significantly as a prognostic marker. It is noteworthy that NM was present at the time of the initial diagnosis in two patients and that pancreatic carcinoma with lymph node involvement was diagnosed after NM in one patient. These patients and others reported in the literature2, 3 suggest that NM can occur before it is possible to identify clinically obvious metastatic disease.

From the review of our data base, the frequency of occurrence of NM from GI malignancies is extremely low. The frequency of NM for patients with any metastatic disease for each primary site was 0.25% of patients with esophageal carcinoma, 0.19% of patients with gastric carcinoma, 0.027% of patients with colon or rectal carcinoma, and 0.023% of patients with and pancreatic carcinoma. These percentages underestimate the true frequencies of occurrence of NM for several reasons. To be included in this study, either imaging or CSF confirmation of the diagnosis was required. We excluded eight patients on this basis. This also would exclude those patients whose physicians considered their overall status so poor that it would obviate further medical intervention. There may have been instances in which the diagnosis was overlooked. Furthermore, many patients who are treated at MDACC and are included in the data base do not receive followup care at MDACC; therefore, the development of NM in those patients is not reflected in the numbers in our study. Notwithstanding these caveats, data from prior studies in gastric carcinoma are remarkably similar to ours. Kim4 reported an incidence of 0.16% and 0.06% for any intracranial metastasis and NM, respectively, from his review of 8080 consecutive patients with advanced gastric carcinoma; and Lee et al.5 reported a prevalence of 0.17% of NM in a review of 11,335 patients with gastric carcinoma.

Among GI primary tumors, NM secondary to gastric carcinoma is reported most commonly.5–22 We were able to identify 60 reports of NM due to gastric carcinoma dating from 1912 through 2004. In 43 of 47 patients (91%), metastatic disease was identified outside the stomach in a location other than the meninges. In 14 instances, the diagnosis was established at autopsy, and no patient identified received antineoplastic therapy for their NM until 1979.6 Table 6 presents information from reports in which treatment was rendered and survival information was available. A recent study by Lee et al.5 reported 19 patients with NM from gastric carcinoma and underscored their poor outcomes; with a median survival of 4 weeks (range, 3–39 wks). Those authors concluded that the use of intrathecal chemotherapy as an independent prognostic factor for survival was encumbered by the obvious selection bias inherent in the initial decision to treat. A group from MDACC reported their experience with NM from gastric carcinoma22 (7 of those patients are included in the current report) and concluded that the incidence of NM had increased in the past several years, although increased detection of NM with better imaging studies may underlie this observation.

Table 6. Literature Review of Patients with Gastrointestinal Carcinoma Treated for Neoplastic Meningitis
StudyHistologyNo. of patientsTreatmentTime (days) from CA dx to NM (mean ± SD)OS (days) from NM dx (mean ± SD)No. with other mets/no. with NM
  • CA: carcinoma; dx: diagnosis; NM: neoplastic meningitis; SD: standard deviation; OS: overall survival; mets: metastases; CS XRT: craniospinal radiotherapy; WBRT: whole brain radiotherapy; IT: intrathecal; Ara-C: cytarabine; MTX: methotrexate; Thio-tepa: triethylene thiophosphoramide; CDDP: cis-diammine dichloroplatinum; FuDR: fluorodeoxyuridine.

  • a

    Systemic death.

Fisher and Weiss, 19796Gastric1CS XRT, systemic chemotherapy60601/1
Baltrucki et al., 198220Gastric1WBRT + IT Ara-C, MTX, and Thio-tepa669111/1
Kim, 19994Gastric5Various128 ± 11643 ± 213/3
Lee et al., 20035Gastric19Various252 ± 28062 ± 8517/19
Current study, 2004Gastric8Various413 ± 55367 ± 867/8
Abdo et al., 200224Esophagus41-IT MTX, CDDP99 ± 7613 ± 8?
Current study, 2004Esophagus7Various418 ± 40278 ± 725/7
Kobayashi et al., 19842Rectal1IT MTX1612??
Smith et al, 198428Colorectal1IT FuDR36571/1
Current study, 2004Colorectal5Various2043 ± 238538 ± 393/5
Ferreira Filho et al., 200133Pancreas1IT MTX, Ara-C12656a1/1
Current study, 2004Pancreas1WBRT, systemic chemotherapy04481/1

NM due to metastatic esophageal carcinoma is rare.23, 24 Only 5 case reports exist in the English language literature (we could only verify 2 of them), and survivals as short as 5 days after diagnosis have been reported.24 Data on extent of disease were unavailable from the literature. In the current report, 5 of 7 patients (71%) had metastatic disease in sites other than the meninges. For the 7 patients presented here who had esophageal carcinoma and NM, the median survival after NM diagnosis was 8 weeks (range, 0–28 wks). Additional information is provided in Table 6.

Colon or rectal carcinoma causing NM also is rare,6, 25–30 with reported survival ranging from 5 days to 5 months after diagnosis. Including the current series, 7 of 9 patients (78%) had metastatic disease in sites other than the meninges. The 5 patients in our series had a median survival of 5 weeks (range, 0–14 wks) after their NM diagnosis. Additional information is in Table 6.

Pancreatic carcinoma with meningeal metastasis is exceedingly rare. We found only seven patients reported in the literature.13, 18, 31–33 To this list, we add an additional patient with pancreatic carcinoma and NM. She survived 64 weeks after the diagnosis of NM (which was diagnosed simultaneously with the primary tumor). In two patients from the literature for whom information was provided, both had non-CNS metastatic disease in addition to NM. Our patient had enlarged, draining lymph nodes. Additional information on a treated patient is presented in Table 6.

We had hoped to identify prognostic indicators in individual patients that could be useful for clinical decision-making, especially because currently available therapies have limited efficacy. Even though the literature reports that patients with NM who present with encephalopathy,34 have CSF flow obstruction on radionuclide flow studies,35 or have bulky CNS disease36 have poorer outcomes, our data do not confirm this, although a larger series may yield more prognostically useful information. Furthermore, although 11 of 21 patients reported herein were diagnosed after 1999 and were treated with modern interventions, their outcomes remained very poor.

Four patients (18%) died during the initial attempt at treatment, a fact that emphasizes the need for rapid diagnosis and intervention. Earlier recognition of NM in patients with GI malignancies has the best chance of having a significant impact on patient outcomes, from both palliative and survival standpoints. Also deserving consideration is whether any intervention at all is warranted for such patients in the face of currently available interventions.

Even with evaluation, diagnosis, and treatment carried out by specialists who are focused on malignant disease and the nervous system, outcomes for patients with NM from GI malignancies have not been impacted significantly. To achieve improved outcomes for these patients, several areas of investigation need elucidation. The identification of the molecular changes that allow for metastases to come to rest and survive in the meninges and CSF will pave the way for targeting or blocking of those factors. Knowledge of these molecular changes may allow for accurate prediction of malignancies with a predilection to metastasize to the meninges, allowing for cost-effective screening for early detection or outright preventive measures. Very little information exists regarding molecular changes that are important for the development of NM, but some factors that may be important include the matrix metalloproteinases,37 urokinase and its soluble receptor,38 and vascular endothelial growth factor.39 Clearly, prevention is more desirable than treatment for this condition. For those patients who nonetheless develop NM, therapies that interfere with cellular machinery that is important for tumor cells to survive and proliferate in the meninges and CSF are needed. Evaluating the activity and pharmacology of antitumor interventions as they relate to the nervous system, its coverings, and the CSF will require continued effort.

Finally, knowledge about survival in treated and untreated patients with NM is important if the efficacy of new therapies is to be evaluated. The current report provides a backdrop of survival data against which future interventions may be compared.


The authors are grateful to Joann Aaron for editorial assistance. The authors 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 data base maintained by the Department of Medical Informatics.