including optic neuropathy (28%), mental status changes (28%), headache (23%), lower extremity weakness (23%), and cerebellar signs (18%) [22].

In a series of 24 patients with LMD due to non-Hodgkin’s lymphoma, 15 patients (63%) presented with cranial nerve palsies, most commonly localized to the facial nerve (10 of 15 patients, or 75%) [19]. Similarly, in another series of 24 patients with non-Hodgkin’s lymphoma, 17 patients (71%) had cranial nerve involvement. In this series as well, the facial nerve was also the most commonly affected cranial nerve, followed by the abducens and oculomotor nerves [20]. Review of seven reported cases of CNS lymphoma with cranial nerve III palsy as the initial presentation of LMD revealed pupil involvement in four patients (57%) [21].

Primary CNS lymphomas appear to behave like systemic non-Hodgkin’s lymphoma with respect to leptomeningeal involvement. In a series of five consecutive patients with T-cell CNS lymphoma and LMD, all five patients presented with symptoms of cranial nerve involvement. Clinical presentations included papilledema and unilateral abducens palsy, bilateral oculomotor palsy, mononeuritis multiplex and hearing loss, bilateral facial nerve palsy, unilateral hearing loss, and the presence of bilateral uveitis, which is usually lacking in systemic lymphomas that present with LMD [23]. A large retrospective series of 45 patients with primary T-cell CNS lymphoma reported by the International Primary CNS Lymphoma Collaborative Group found one patient with exclusive leptomeningeal involvement presenting with cranial nerve palsies but no obvious intracranial lesions [24]. The same happens with the more common B-cell primary CNS lymphoma.

32.3.3 LMD due to Primary Brain Tumors

LMD arising from primary brain tumors is rare and usually occurs late in the course of the disease. In a review of 52 patients with malignant gliomas, 11 patients (21%) were found to have leptomeningeal involvement at autopsy. Of these, ten patients had hydrocephalus, nine patients had tumor invasion of the lateral ventricles, eight patients had spinal subarachnoid seeding with involvement of the cauda equina, and three patients had symptomatic compression at the thoracic or lumbar level [25]. In general, gliomas have a predilection for the ventricles and may drop along the CNS to involve the spinal nerve roots. Cranial nerve lesions occur less frequently than in solid or hematogenous tumors. For glioblastoma multiforme, LMD manifestations have included optic neuropathy and oculomotor, abducens, and facial nerve palsies; there have been no reports in the literature of trochlear nerve palsies [25–27].

32.4 Diagnosis

The fact that autopsy studies consistently show incidences of LMD exceeding those in clinical series indicates that many cases of LMD go undiagnosed. The Memorial Sloan-Kettering Cancer Center review of 2375 autopsies of patients with cancer found an 8% incidence of LMD, compared to 5% reported in most clinical studies [12, 13].

Similarly, the National Cancer Institute study of small cell lung cancer showed an 11% incidence of LMD antemortem compared to 25% postmortem [14].

Multiple factors contribute to the difficulty of early diagnosis of LMD, including nonlocalizing signs and symptoms, which are frequently ascribed to the primary tumor or to a side effect of the medical treatment. Also, cerebrospinal fluid (CSF) analysis may be challenging. In a study by Glass et al. [28], 41% of patients with negative CSF cytology antemortem were found to have LMD at autopsy. This finding was supported by Wasserstrom et al. [9] and Murray et al. [29], who independently demonstrated variable yields of malignant cells at different levels in the CNS, with the highest concentrations localizing to the cisternal and ventricular compartments. More specifically, Wasserstrom et al. [9] identified a subgroup of approximately 5% of patients with negative lumbar punctures who had positive cytology findings only from CSF derived from the ventricles or cisterna magna. Another subgroup of approximately 9% of patients had consistently negative CSF cytology findings despite radiographically unequivocal disease [9]. It follows that the optimal number of samples necessary to diagnose patients with LMD by cytology has not been established. While some authors have noted a significant increase in sensitivity with repeat testing in initially cytologically negative cases [2, 10, 30], others have reported no added diagnostic benefit from conducting more than two lumbar punctures [4, 9]. Of interest is that despite the high rate of false-negative cytology findings (up to 50% in some series), a completely normal lumbar puncture is associated with LMD in fewer than 5% of cases [9, 31]. Abnormalities in CSF on initial lumbar puncture in patients with LMD have been reported to include a high opening pressure (>160 mmH2O; 50% of lumbar punctures), high protein level (>50 mg/dl; 81%), hypoglycorrachia (<40 mg/dl; 31%), and pleocytosis (>5 white blood cells/mm3; 57%) [9].

Although a positive CSF cytology finding is the diagnostic gold standard for LMD [30], other ancillary biochemical CSF markers have been described in the literature, in particular for LMD due to solid and hematogenous tumors. These include carcinoembryonic antigen, α-fetoprotein, gastrin-releasing peptide, β- human chorionic gonadotropin, β2-microglobulin, β-glucuronidase, soluble CD27, LDH isoenzyme-5, glucose-6-phosphate isomerase, and various types of monoclonality. The latter include immunoglobulin gene rearrangements identified by flow cytometry or polymerase chain reaction. In patients with negative CSF cytology findings, positive surface markers support the diagnosis of LMD. These ancillary markers may also be used to monitor treatment response [12, 32–38].

32.4.1 Radiographic Imaging

Radiographic modalities useful in the diagnosis and staging of LMD include cranialenhanced computed tomography (CT), CT myelography, brain and spine magnetic resonance imaging (MRI), and radionuclide CSF flow studies. Cranial-enhanced CT is abnormal in approximately 25% of patients with LMD. The abnormalities observed include parenchymal volume loss, ependymal or subependymal

enhancement, subarachnoid enhancing nodules (less common in hematologic malignancies), sulcal–cisternal enhancement or obliteration, intraventricular enhancing nodules, communicating hydrocephalus, and irregular tentorial enhancement [4]. Brain MRI has a greater sensitivity (up to 76%) than cranial-enhanced CT but lower reported sensitivity for hematologic tumors (55%) than for solid tumors (90%) [16, 39]. Brain MRI is the imaging modality of choice for the diagnosis of intracranial LMD. Findings on brain MRI in patients with LMD include leptomeningeal, subependymal, dural, or cranial nerve enhancement. Table 32.1 outlines the differential diagnosis of cranial nerve enhancement and Fig. 32.1 shows examples of MRI findings in LMD, superficial cerebral lesions, and communicating hydrocephalus [40]. While CT and MRI are more useful for identifying bulky CNS disease, CSF flow studies provide information regarding the functional anatomy of CSF spaces. This allows for targeted treatment of areas with flow abnormalities. In addition, functional CSF status as determined by CSF flow studies has been correlated with patient survival [4, 7, 31, 41–44].

32.4.2 Optic Neuropathies in LMD

An isolated optic neuropathy may be the presenting sign of LMD in cancer patients. Among patients with leptomeningeal optic nerve involvement, up to 44% experience

Table 32.1 Differential diagnosis of cranial nerve enhancement [42]

Fig. 32.1 Cranial nerve involvement in four cases of LMD. Contrast-enhanced axial T1weighted images show enhancement and involvement of cranial nerves: (a) left optic nerve sheath (arrows); (b) oculomotor nerves (arrows); (c) left Meckel’s cave (arrows); and (d) right facial nerve (arrows)

visual loss [43]. However, patients may have normal vision at the time of their initial assessment. An almost constant finding in these cases is a relative afferent pupillary defect. The magnitude of this defect as measured by neutral density filters may be used to clinically follow these patients. Color vision is likewise affected early, resulting in confusion of the red–green axis on Hardy–Rand–Ritter or Ishihara color plate testing. Visual field changes may include a central scotoma, diffuse depression, or nonspecific visual field changes. Dilated fundus examination may reveal no abnormality; it may also reveal optic disc edema due to raised intracranial pressure and/or optic nerve infiltration, optic disc hemorrhages, optic disc pallor or atrophy, retinal or vitreous hemorrhages, or engorged retinal vasculature. A diagnostic quartet of leptomeningeal optic nerve infiltration has been proposed by McFadzean et al. [44] involving the combination of headaches typical of raised intracranial pressure, blindness, sluggish or absent pupillary reflexes, and normal-appearing optic discs. Although long-standing papilledema may lead to post-stasis optic atrophy, optic nerve damage usually results from infiltration of the tumor in the optic nerve sheath. This may cause a compartment syndrome resulting in posterior segment ischemia. In this situation, patients may experience rapid and profound visual loss