Cerebrospinal fluid pleocytosis

Pitfalls and benefits of combined analysis using cytomorphology and flow cytometry


  • Martin Bommer MD,

    Corresponding author
    1. Department of Internal Medicine III, (Hematology, Oncology, Rheumatology and Infectious Diseases), University Hospital of Ulm, Ulm, Germany
    • Department of Internal Medicine III—Hematology, Oncology, Rheumatology and Infectious Diseases, University of Ulm, Albert-Einstein-Allee 23, D-89081 Ulm, Germany
    Search for more papers by this author
    • Fax: (011) 49-731-500-45515

  • Andreas Nagy MD,

    1. Department of Internal Medicine III, (Hematology, Oncology, Rheumatology and Infectious Diseases), University Hospital of Ulm, Ulm, Germany
    Search for more papers by this author
  • Christine Schöpflin,

    1. Department of Internal Medicine III, (Hematology, Oncology, Rheumatology and Infectious Diseases), University Hospital of Ulm, Ulm, Germany
    Search for more papers by this author
  • Sandra Pauls MD,

    1. Department of Radiology and Nuclear Medicine, Sana Hospitals Lübeck, Lübeck, Germany
    Search for more papers by this author
  • Mark Ringhoffer MD,

    1. Medical Department III, Municipal Hospital Karlsruhe, Germany
    Search for more papers by this author
  • Mathias Schmid MD

    1. Department of Internal Medicine III, (Hematology, Oncology, Rheumatology and Infectious Diseases), University Hospital of Ulm, Ulm, Germany
    Search for more papers by this author



Cerebrospinal fluid samples with doubtful morphologic interpretation are a common problem in the workup of patients with clinical signs for leptomeningeal disease. The authors report on the combination of morphology and flow cytometry in the diagnosis of leptomeningeal disease in patients with radiological, clinical, or cytological findings suspicious for leukemia or lymphoma with spread into the cerebrospinal fluid.


The authors defined a set of antibodies for flow cytometric analysis, which is capable of distinguishing between malignant and nonmalignant hematopoietic cells. One hundred twenty-seven cases were analyzed with both methods.


The additional application of flow cytometry resulted in an improvement of diagnostic reliability in 29 of 127 cases. Diagnostic sensitivity was raised from 73% (cytology) to 96% (flow cytometry), specificity from 94% to 97%, the positive predictive value from 88% to 96%, and the negative predictive value from 76% to 97%.


Because the appropriate selection of markers is crucial to successful analysis, the authors suggested a highly selected panel of antibodies for flow cytometry analysis of cerebrospinal fluid samples. The authors were able to demonstrate that leptomeningeal disease caused by leukemia or lymphoma can be diagnosed by flow cytometry and discriminated from reactive pleocytosis in most cases of doubtful morphology. Cancer (Cancer Cytopathol) 2011. © 2011 American Cancer Society.

Leptomeningeal spread of tumor cells is a rare, but in many cases, it is a fatal event with an incidence ranging from 5% in solid tumors up to 20% and more in leukemia and lymphoma.1, 2 In the past decade, we observed an increasing importance of this complication due to successful treatment of systemic disease even in relapse. Diagnosis of neoplastic meningitis (NM) is obtained using magnetic resonance imaging (MRI) and analysis of the cerebrospinal fluid. Using cerebrospinal fluid (CSF) cytology is of limited sensitivity, approximately 50%-60%, but of excellent specificity (>95%) when performed by an experienced cytologist.3 Nevertheless, CSF analysis including cytology may lead to doubtful or false results in up to 40% of samples.4 This is often caused by the paucity of tumor cells and the presence of reactive lymphocytes—mostly of T-cell origin—making it awkward to distinguish malignant hematopoietic blasts by morphology alone. For almost 25 years, flow cytometry (FC) has been routinely used for the diagnosis and further subtyping of leukemia and lymphoma in blood, bone marrow, and body fluid samples.5 It is an objective method for qualitative and quantitative analysis of cell suspensions and can detect small populations of tumor cells with aberrant surface-marker expression through multicolor and multiparameter analysis.6 During the past 10 years, several reports have been published with encouraging results on the diagnostic value of flow cytometry of CSF samples from patients with leukemia and lymphoma.7 Flow cytometry can probably increase the sensitivity of detection of tumor cells in patients with aggressive lymphoma at high risk for central nervous system (CNS) involvement8, 9 or in patients with atypical or suspicious CSF morphology.10



Between 1996 and 2008, we performed both cytology and flow cytometry of CSF samples from 127 patients; 96 of these 127 patients had previously been diagnosed with leukemia or lymphoma. Thirty-one patients had no known underlying malignant disease at the time of lumbar puncture, but clinical data suggested malignant leptomeningeal disease. The final clinical diagnosis was established by using all data available including reports from microbiology, virology, and pathology during the clinical follow-up of the patients. We did not perform biopsies to obtain histopathology of the meninges.

All CSF samples for flow cytometry were drawn by lumbar puncture during standard clinical workup. All patients gave written informed consent for the procedure. There were 60 women and 67 men with a median age of 53 years ranging from 16 to 75 years. All data were taken from our clinical database and the patients' charts. This retrospective study was conducted in accordance with the principles of the Helsinki Declaration. The local ethics committee approved the evaluation protocol.

CSF analysis

Cerebrospinal fluid (CSF) was obtained by lumbar puncture. Precautions were taken to minimize contamination by peripheral blood.

The fresh specimens were processed following standard procedures previously published.11 All smears were categorized by one of us (M. B.) to be malignant, suspicious, benign, or atypical. The term “malignant” was used only in cases with clear morphologic changes consistent with lymphoma, leukemia, or carcinoma. The term “suspicious” was used for samples with cells that did not meet all criteria for malignancy (see Fig. 1). Samples with elevated cell counts consisting of small lymphocytes, macrophages, and ependymal cells were judged to be reactive pleocytosis and classified as “benign”. All samples that could not be classified unambiguously were designated “atypical”. Samples with both atypical and suspicious morphology were considered to be of doubtful morphology. All samples with “malignant” or “suspicious” morphology were classified as “positive” (+). All samples with “benign” or “atypical” morphology were classified as “negative” (−). FC analysis was done independently within a few hours of sampling. CSF volume ranged from 1.5 mL to 6 mL. The fresh specimens for FC were centrifuged, resuspended in phosphate buffered saline (PBS), and diluted according to cell count. Antibodies were added according to the manufacturer's instructions. The antibody panels were selected according to the number of cells and the previously known expression profile either detected by histology or flow cytometry data from blood or bone marrow specimens (see Table 1). All samples from patients without a known underlying malignant disease were assessed with the B-cell lymphoma panel to differentiate between T- and B-cells (see Table 1). Because of low cell counts, not all specimens could be assessed with the full marker panel. The 3- or 4-color flow cytometry was performed from 1996 until May 2000 with a BD FACScan (Becton Dickinson, Franklin Lakes, NJ). Since May 2000, we used an EPICS-XL-MCL (Beckman Coulter, Fullerton, Calif) with Expo32 Software (Beckman Coulter, Fullerton, Calif). Cell populations were gated on forward scatter (FSC) and side scatter (SSC), CD 19 or CD20, CD 3, and CD 45 depending on the underlying or suspected diagnosis.

Figure 1.

Cerebrospinal fluid sample of a patient suffering from malignant lymphoma (ID 64) was judged to be of doubtful morphology (“suspicious”). Note the large lymphocytes with “blastic appearance” but without all criteria of malignancy. Finally, flow cytometry revealed clonal B-cell population. The photograph was taken using an Axioskop 50 microscope with oil immersion objective Epiplan NeoFluar 100x and an AxioCam MRc5 digital camera (all from Carl Zeiss MicroImaging GmbH, Göttingen, Germany).

Table 1. Marker Profiles Used Depending From Underlying or Suspected Diagnosis
 Myeloid LeukemiaLymphoproliferative Disorder
  1. NK indicates natural killer cell lymphoma; TdT, terminal deoxynucleotidyl transferase.

Basic panelCD45CD45CD45CD45CD45
Supplementary panelMPOCD38CD1aKappaCD2
 CD13CD5CD56 CD8

The results were analyzed by 2 physicians (M. S. and coworkers). Samples were judged to be malignant only when a distinct cell population with the disease-specific immunophenotype could be detected.

Statistics were performed using SPSS (IBM SPSS, Chicago, Ill) software.


Overall population

From each of the 127 patients, 1 sample was analyzed. Final judgment resulted in 52/127 patients to have leptomeningeal spread of malignancy, 50/52 patients had a history of hematopoietic malignancy. Two patients were found to have nonhematopoietic cancer (breast cancer, lung cancer) with unusual morphology of the cerebrospinal fluid. Seventy-five of 127 patients were found to have no cerebrospinal involvement despite a history of hematopoietic malignancy in 46 of 75 cases. Using cytomorphology, we judged 38 samples to be malignant, 29 samples were of doubtful morphology (“atypical” or “suspicious”), and 60 samples were judged to be benign. Flow cytometry revealed 52 samples to be malignant and 75 samples to be benign. Compared with the final diagnosis, cytological analysis resulted in 4 false-positive and 3 false-negative judgments. Flow cytometry yielded 2 false-negative and 2 false-positive results. The overall average cell count was 220/μL. The average cell count of malignant samples was 404/μL and of benign samples 96/μL. For a comprehensive overview of the results see Table 2.

Table 2. Summary Including Patient Characteristics and Results
Patient CharacteristicsTotalLeptomeningeal Malignancy FoundBenign PleocytosisCytologyFlow CytometryCombined Analysis
 + + + 
  • MAL indicates malignant; FP, false positive; SUS, suspicious; BEN, benign; ATY, atypical; FN, false negative; PPV, positive predictive value; NPV, negative predictive value; SPE, specificity; SENS, sensitivity.

  • a

    Cytology vs FC (McNemar-Test).

  • b

    Cytology vs combined analysis (McNemar-Test).

Underlying malignancy991625253%404 4730 381243713390%73%88%72%52247296%96%94%96%52247096%100%100%100%
 Known at index puncture961865052%420 4630 361243612390%72%88%72%52244096%98%98%96%500460100%100%100%100%
   ALL14356750%656 77 510801100%71%86%100%7070100%100%100%100%7070100%100%100%100%
   AML12170975%201 375 620311100%67%89%67%9030100%100%100%100%9030100%100%100%100%
   DLBCL401241845%263 2215 1354157179%72%78%64%180220100%100%100%100%180220100%100%100%100%
   Indolent B-cell   lymphoma234321043%887 1356 910940100%90%100%69%12211085%100%100%100%10213087%100%100%100%
   T-cell lymphoma671583%85 13 320100-50105010
   NK1801 80 0 01000010001000
Not known at index puncture320267%24 111 21322010
  Carcinoma3 2   1  21322010
No underlying malignancy detected28207 28207 23582%2828
 Viral menigitis/ encephalitis10302 10302 821010
 Bacterial menigitis/ encephalitis7183 7183 5277
 Noninfectious Inflammatory disease11128 11128 1011111

Subgroup analysis

History of myeloid disease (AML, blast crisis of CML, and CMML)

We analyzed CSF samples from 12 patients with myeloid disease including acute myeloid leukemia (AML), blast crisis of chronic myeloid leukemia (CML, BC), and chronic myelomonocytic leukemia (CMML). Average cell count was 170/μL with a range from 2/μL to 773/μL. Nine samples were malignant, 6 of them correctly diagnosed by cytology. Two samples were suspicious, and 1 sample was atypical by morphology, but all 3 samples were found to be malignant when FC was used (ID 32, ID 33, ID 36). The 3 remaining samples were identified by both methods to be benign despite elevated cell counts. Those 3 samples showed pleocytosis dominated by T-lymphocytes and mononuclear cells lacking AML-specific markers.

History of acute lymphoblastic leukemia (ALL)

Fourteen samples from patients with ALL were processed: c-ALL (n = 9), B-ALL/Burkitt-type (n = 2), and T-ALL (n = 3). Average cell count was 356/μL with a range from 1/μL to 2233/μL. Seven samples were malignant, 5 of them correctly diagnosed by morphology, and 1 sample was suspicious. One sample judged as benign by morphology was found to be malignant by flow cytometry (ID 65); despite low cell count a distinct population of >40% expressing CD10 and CD19 (as previously found in the bone marrow) could be detected.

History of Non-Hodgkin lymphoma

Seventy samples from patients with non-Hodgkin lymphoma were analyzed. Of these, 40 patients suffered from diffuse large B-cell lymphoma (DLBCL), 6 patients from T-cell lymphoma (T-NHL), 23 patients from indolent lymphoma (4 multiple myeloma [MM], 4 mantle cell lymphoma [MCL], 11 chronic lymphocytic leukemia [B-CLL]), 4 patients from B-cell lymphoma not otherwise specified (B-NHL NOS), and 1 patient from NK-cell lymphoma. Average cell count was 217/μL with a range from 1/μL to 5000/μL. Finally, 34 samples were classified as malignant.

Thirty-four samples were classified by morphology to be malignant (n = 25) or suspicious (n = 9). Four samples from patients with DLBCL were false positive by cytology and judged to be benign by flow cytometry (ID 91, 94, 100, 109).

Of the thirty-six samples classified by morphology as benign (n = 25) or atypical (n = 11), confirmation by FC was found in 34 cases. Three samples with atypical cytology were stated as reactive by FC (ID 67, 121) or malignant (ID 68).

Two patients with known B-CLL and unexplained neurologic symptoms suspicious of CNS involvement by lymphoma were found to have pleocytosis dominated by lymphocytes; in both cases, cytology was categorized as “benign”. Flow cytometry analysis revealed in both cases a clonal B-cell population with 32% (ID 80) and 2% (ID 88) of all nucleated cells. The case with ID 80 was interpreted as leptomeningeal involvement, but VZV myelitis diagnosed a few days later made this interpretation doubtful. In the case with ID 88, spontaneous regression of neurologic symptoms despite lack of a specific antineoplastic treatment along with a low percentage of clonal B-cells indicated a reactive pleocytosis as suggested by morphology.

No history of malignant disease

Thirty-one patients underwent CSF analysis because of clinical or radiological changes suspicious for neoplastic meningitis without any known pre-existing malignancy at the time of lumbar puncture. Twenty-nine patients had no NM. Clinical diagnosis included acquired immune deficiency syndrome (AIDS), viral encephalitis, brain abscess, tuberculosis, sarcoidosis, pachymeningeosis, and carcinoma of unknown primary (CUP). Average cell count was 188/μL with a range from 8/μL to 840/μL. The average cell count of the malignant samples was 24/μL and for nonmalignant samples, 200/μL. Twenty-three samples were judged morphologically to be benign; 5 were atypical. In all 29 samples, the nucleated cells exhibited predominantly T-cell markers without aberrant marker profile and low or absent B-cells.

In 2 patients (ID 1, ID 2), neoplastic meningitis due to nonhematopoietic tumor (breast cancer, adenocarcinoma of the lung) was found. Both samples were unambiguously assessed by morphology as leptomeningeal infiltration by nonhematopoietic malignancy, whereas FC showed high T-cell count without aberrant marker expression and low or absent B-cell proportion. Because we did not test for epithelial markers, these 2 samples were judged by FC to be reactive.


Assessing CSF samples by cytomorphology is often limited by low tumor load, ambiguous morphologic appearance, and bloody contamination. Particularly in patients with leukemia, lymphoma, and CNS infection, reactive lymphocytes are sometimes difficult to distinguish from blasts. Due to the finding that reactive pleocytosis is mainly caused by inflammatory T-cells, NM can be ruled out by demonstrating a T-cell dominant reactive cellular pattern without cells expressing malignancy specific markers. Because FC analysis is able to detect small cell populations with a distinct marker profile, its sensitivity is reported to be excellent.9 On the other hand in cases with circulating tumor cells (eg, in indolent lymphoma) and complicating inflammatory disease, misinterpretation is triggered by failure of the blood-brain barrier with penetration of blood cells into the CSF without real infiltration of the CNS. This phenomenon in the context of inflammatory diseases has already been described elsewhere.12 FC analysis is, as morphology, not able to distinguish between real leptomeningeal malignancy and “pseudomeningeosis” caused by inflammatory penetration or contamination during lumbar puncture. On the other hand, reactive pleocytosis with “blastic” appearance in patients without known underlying malignant disease can easily be discriminated from hematopoietic malignancy by using FCM.

Because of the limited amount of CSF fluid in routine clinical situations, it is necessary to define a distinct set of markers. Based on our experience, we suggest a basic pattern of antibodies (see Table 1) that enable the investigator to categorize the majority of cells occurring in CSF. Expression of CD45 allows the differentiation of hematopoietic from nonhematopoietic cells (except in rare cases of erythroid malignant disorders that may lack CD45 expression). CD3 and CD19 provide information about the B:T-ratio. CD19 or CD20 as markers specific for B-cells are to be combined with Kappa/Lambda to prove B-cell clonality in all major subtypes of B-NHL. Expression of early markers such as CD10 and TdT are mandatory in detecting early B-precursors such as c-ALL. To detect T-cell lymphoma, marker selection has to be more meticulous: most of the T-cell markers are expressed on normal mature T-cells (CD3, CD4, CD8). Only aberrant marker coexpression proves malignancy in those cases. In general, if an underlying disease is recognized, marker selection should be led by the known marker expression profile. It is not our approach to detect NM of solid tumors by flow cytometry, but the unusual “blastic” morphology of the cells analyzed (ID 1, ID 2) triggered FC analysis. Our marker panel adds little information on epithelial cells; it enabled us to distinguish poorly differentiated nonhematopoietic tumors from hematopoietic malignancy (eg, by lack of CD45, CD 19, CD 2).

The results of previous authors and our findings raise the important question about the “gold standard” of CSF analysis in hematopoietic malignancies. Up to now, morphology is the standard procedure. Because most studies, including the present one, were performed with symptomatic patients, this is perhaps only true for a high tumor burden causing symptoms. Studies performed with asymptomatic patients but with a high risk for leptomeningeal disease show different results: low detection rates using morphology in aggressive lymphoma are correlated with a low tumor load but can be enhanced by the use of flow cytometry.9 The population of patients treated has changed during the past years. Many patients with relapsed or refractory leukemia and lymphoma are treated with curative intent by using autologous or allogeneic stem cell transplantation. Advanced disease is associated with increased risk for NM. This is particularly true for solid tumors and NHL13, 14 and seems to be of increasing importance in leukemia.11 Because CNS-targeted therapies like radiation, high-dose chemotherapy, and intrathecal therapy remain associated with significant long-term toxicity, reliable tests to identify occult leptomeningeal disease in leukemia and lymphoma in at-risk asymptomatic patients are required. We propose the use of cytomorphology combined with FC to optimize diagnostic efficacy.


Flow cytometry is able to enhance the diagnostic impact of CSF examination in patients with suspected CNS involvement by hematopoietic malignancies. Using distinct marker panels according to the known or suspected malignancy, flow cytometry combined with conventional morphology is helpful and reliable. Especially in situations with doubtful morphology, flow cytometry gives additional information to discriminate reactive from malignant pleocytosis.


The authors made no disclosures.