Examination of Cerebrospinal Fluid

Definitive diagnosis of NM is made by identification of malignant cells in CSF. The relative sensitivity of this test depends on the number and volume of CSF specimens examined, the state of preservation of the CSF cells, and the source of the specimen (lumbar or cisternal being more reliable than intraventricular fluid). In one study, the sensitivity of the initial CSF examination was approximately 50% to 60% and improved to about 80% with a second specimen, and then each subsequent sample increased sensitivity by 2% to 5%.¹³ In another study, the yield of a single CSF analysis improved if the CSF volume studied exceeded 10.5 mL and was removed from the location (lumbar, cisternal, ventricular) nearest the symptomatic site or the area of greatest involvement, as seen on neuroimaging studies.¹⁴ Most cytopathologists prefer to assess serial samples obtained on different dates rather than a single large-volume specimen.

Although nonspecific, CSF protein determination has historically been the most sensitive indicator of NM because this protein level is abnormal (>45 mg/dL) in approximately 80% to 90% of patients.¹³ Conversely, finding a normal CSF protein reading is relatively strong (but not absolute) evidence against this diagnosis. Lumbar CSF is more likely than ventricular fluid to have elevated protein and malignant cells. Cisternal fluid has also been proposed to be a more reliable indicator, but its acquisition may carry greater risk.¹⁵ If high elevations (>500 mg/dL) of CSF protein are found in lumbar fluid, either NM is advanced or there is a partial or complete blockage of CSF flow from cephalad locations. Approximately 30% to 57% of patients have lumbar CSF pressure higher than 150 mm H₂O, and approximately 31% to 55% have a reduction in CSF glucose levels. In patients with subsequently verified NM, less than 5% have completely normal CSF profiles.⁸

The usefulness of newer techniques such as fluorescent in situ hybridization (FISH) for specific gene amplifications or deletions, flow cytometry, and DNA fragment amplification by polymerase chain reaction (PCR) is currently being investigated. Determination of T-cell receptor and immunoglobulin heavy-chain gene rearrangements by PCR using DNA amplified from CSF lymphocytes can be helpful when evaluating a patient for suspected lymphomatous meningitis, particularly if the gene rearrangements in the systemic lymphoma tissue are known. The systemic tissue source can occasionally be implied by testing for soluble CSF tumor-associated antigens (e.g., carcinoembryonic antigen, CA-125, human chorionic gonadotropin, α-fetoprotein) or with the use of immunocytochemistry (e.g., HMB-45 for melanoma and OCT-3-4 for germ cell tumors).

Neuroimaging

Currently, magnetic resonance imaging (MRI) of the brain and spine with gadolinium contrast enhancement is the procedure of choice. Findings on contrast-enhanced MRI will appear abnormal during the course of the disease in approximately 70% of patients with NM.¹⁶ High-resolution MRI is necessary for proper evaluation of the meninges of the spine. In general, lower resolution “open” MRI is less reliable for identification of NM unless florid abnormalities are present. Patchy nerve root enhancement, matting of nerve roots of the cauda equina, nodular deposits, and linear, continuous enhancement of the pia-arachnoid of the conus medullaris, brainstem, and cerebrum are hallmarks of the diagnosis (Fig. 137-1). Patchy, asymmetric enhancement of the leptomeninges over the cerebral convexities and in the sulci, basilar cisterns, insulae, pituitary stalk, cranial nerve roots, or superior cerebellar folia (or any combination of these sites) may be observed (Fig. 137-2). Indirect signs include communicating hydrocephalus, bilateral transependymal edema, and effacement of convexity sulci. Contrast enhancement of the meninges can be associated with many non-neoplastic conditions, including prior or current infections, previous subarachnoid hemorrhage, previous lumbar puncture or intrathecal chemotherapy, prior neurosurgical procedures (e.g., placement of an intraventricular reservoir), chronic CSF leak syndrome, and other neurological conditions.¹⁶ Nevertheless, in a patient with a high KPS score and active systemic cancer in whom symptoms or signs and neuroimaging findings strongly suggest NM, some physicians believe that treatment is justifiable even if the CSF cytologic study is negative.^17,18

FIGURE 137-1 A, Patchy nerve root enhancement, matting of nerve roots of the cauda equina, and nodular deposits (arrows) in a patient with acute lymphocytic leukemia. B, Linear, continuous enhancement of the pia-arachnoid surrounding the conus medullaris (arrows) in a patient with metastatic melanoma.

FIGURE 137-2 Marked enhancement of the leptomeninges and cerebral sulci and within the superficial folia of the superior cerebellum (arrows).

Cerebrospinal Fluid Flow Studies

Some investigators have proposed that CSF flow dynamics should be evaluated before initiating treatment of NM. CSF flow is usually assessed by nuclear medicine techniques with either ¹¹¹In-diethylenetriamine pentaacetic acid or, less commonly, ⁹⁹Tc-labeled albumin. Partial or total blockage of CSF flow has been identified in up to 30% to 70% of patients with NM.^19,20 The stated rationale for performing flow studies includes identifying CSF flow blockage before treatment, preventing accumulation of high concentrations of administered drug in areas of CSF loculation, and conversely, identifying areas that would not receive adequate drug concentration beyond the areas of blockage. Retrospective studies have suggested that outcome is poorer in patients with CSF blocks or those in whom blocks cannot be opened with focal radiation therapy (RT).^19,20 To date, routine use of flow studies has not gained widespread acceptance among practitioners.

Meningeal Biopsy

Occasional patients are appropriate candidates for meningeal biopsy, but most are not because of poor performance status and comorbid conditions. The best candidates are patients with high clinical suspicion of NM but whose CSF cytologic studies are repetitively negative. Nevertheless, the sensitivity and specificity of this technique have proved to be low, particularly when there is no acceptable target showing MRI contrast enhancement, and the procedure carries added risks for the patient.²¹

Surgery

Neurosurgical procedures may include placement of intraventricular (e.g., Ommaya) reservoirs for CSF access, ventriculoperitoneal shunting for hydrocephalus, and the occasional meningeal biopsy. Justifications stated for Ommaya placement include the ease of repetitive access to CSF, decreased patient discomfort in comparison to lumbar access, and improved drug distribution within the CSF pathways after intraventricular administration.²³ Except for patients with childhood ALL, clinical studies have not shown statistically significant improvement in overall survival when comparing intraventricular with intralumbar drug administration. The intraventricular reservoir is typically placed below the galea in the right frontal region (in left hemisphere–dominant individuals), with the catheter located within the right frontal horn of the lateral ventricle.²⁴ Complications of reservoir placement occur in approximately 10% of patients and include aseptic meningitis (43%), malpositioning (2% to 12%), obstruction of the catheter (6%), and bacterial infection (2% to 13%). Bacteria isolated from infected systems are usually Staphylococcus epidermidis or Propionibacterium, although more pathogenic organisms, including Staphylococcus aureus, Streptococcus pyogenes, and gram-negative species, can be found.²⁵

There has been some controversy about placement of ventriculoperitoneal shunts in patients with high intracranial pressure accompanying NM. Major concerns have included the questionable benefit of this procedure in patients with end-stage disease, the potential systemic dissemination of malignant CSF cells to the abdomen, ineffective drug delivery to CSF pathways because of siphoning of administered drug from the intraventricular reservoir, and other medical complications (e.g., infection, CNS hemorrhage) associated with placement of these devices. In an effort to theoretically limit drug egress, some physicians have installed in-line on-off valves and, recently, programmable valves that can be set to temporarily impede shunt function. This allows the system to be functionally closed for an interval that is estimated from the patient’s known shunt dependency status. However, these manipulations carry significant risk and should be performed only by those with considerable experience in these specific techniques. It should be noted that temporary closure of shunt valves has not yet been shown in clinical studies to sustain therapeutic levels of a drug in CSF or to improve survival.

Radiation Therapy

Craniospinal irradiation (CSI) has been part of the management of patients with medulloblastoma or pineoblastoma who are considered poor risks for surgery, and it has also been used for salvage therapy in patients for whom initial therapy for leukemic meningitis has failed. Because CSI encompasses much of the bone marrow, the resulting hematologic toxicity may affect the ability to provide substantial cytotoxic chemotherapy in future regimens. In NM from solid tumors, CSI has not been proved to provide a survival benefit in controlled studies. However, focal RT delivered to areas of bulky disease or symptomatic sites often provides palliative benefit. Some investigators are concerned that the toxicity of intrathecal or high-dose systemic methotrexate (MTX) administration may be increased after whole-brain RT and note that the increase is somewhat less likely if MTX is administered before RT. The relationship of drug toxicity to RT is less well understood for the available intrathecal agents other than MTX.

Chemotherapy

Many patients in whom NM is diagnosed are too ill to receive aggressive therapy, and they opt for supportive or hospice care. However, there remain certain younger patients with high KPS scores and controlled systemic disease who many believe are more appropriate candidates for drug therapy. Many systemically administered agents exhibit incomplete penetration of the blood-brain barrier. Penetration (CSF concentration/systemic blood concentration) is reasonable for thiotepa (90%), topotecan (30%), temozolomide (30%), and cytarabine (ara-C; 20% to 28%) but low for MTX (3%).²⁶ Some response to systemically administered capecitabine has been reported in patients with NM from breast adenocarcinoma.^27,28 Temozolomide has produced responses in NM accompanying malignant gliomas.²⁹ In a small study of 31 patients, cytologic clearing of tumor cells from CSF was as frequent with systemic administration of high-dose MTX as with intrathecal administration of the conventional dose. Survival time was also significantly longer (13.8 versus 2.3 months, P = .003) in patients receiving systemic MTX than intrathecal MTX.³⁰ In another small study, 104 patients were randomized to receive either RT plus systemic chemotherapy (various agents) and intrathecal chemotherapy or RT plus systemic chemotherapy alone. There were no differences in outcome, with both groups having a median overall survival interval of 4 months, and more toxicity was observed in the patients receiving intrathecal chemotherapy.³¹

Theoretically, intrathecal administration can provide adequate CSF drug delivery and thus obviate the need for the high, potentially toxic systemic doses necessary to achieve the same concentrations. However, for most neoplasms, survival has not yet been shown to be superior after intrathecal treatment. There are several potential barriers to intrathecal treatment with respect to chemotherapeutic agents, including their short half-life, cell cycle specificity, CSF compartmentalization, and inadequate penetration of the CNS parenchymal surfaces. Previous studies have reported that only 55% of CSF tumor cells cycle in a 10-day span, with most cells in G₀, only 0.1% in S phase, and 1% in mitosis.²³ In contrast, the half-life of most intrathecal agents is measured in only minutes or a few hours at most.³² Some agents are not converted to active metabolites within CSF; for example, triethylenephosphoramide (TEPA), the active metabolite of thiotepa, is not measurable in CSF after intrathecal administration.

Nevertheless, agents that have been administered intrathecally, either in anecdotal reports or in clinical trials, include MTX, ara-C, sustained-release ara-C (depo-ara-C), thiotepa, mafosfamide, etoposide, rituximab, interferon alfa, and topotecan. The first four of these agents have been used most often in clinical practice. It should be stressed that many of these agents have not yet been approved by the Food and Drug Administration (FDA) for this specific indication. Depo-ara-C has received FDA approval for the treatment of lymphomatous meningitis, and MTX and ara-C are indicated for the treatment of lymphomatous and leukemic meningitis. The results of phase II studies and randomized controlled clinical trials involving intrathecal agents are summarized in Tables 137-1 and 137-2.