Detection of malignant hematopoietic cells in cerebral spinal fluid previously diagnosed as atypical or suspicious


  • Malcolm Schinstine Ph.D., M.D.,

    1. Cytopathology Section, Laboratory of Pathology, National Cancer Institute, National Institutes of Health, Bethesda, Maryland
    Search for more papers by this author
  • Armando C. Filie M.D.,

    1. Cytopathology Section, Laboratory of Pathology, National Cancer Institute, National Institutes of Health, Bethesda, Maryland
    Search for more papers by this author
  • Wyndham Wilson M.D., Ph.D.,

    1. Experimental Transplantation and Immunology Branch, National Cancer Institute, National Institutes of Health, Bethesda, Maryland
    Search for more papers by this author
  • Maryalice Stetler-Stevenson M.D., Ph.D.,

    1. Cytopathology Section, Laboratory of Pathology, National Cancer Institute, National Institutes of Health, Bethesda, Maryland
    Search for more papers by this author
  • Andrea Abati M.D.

    Corresponding author
    1. Cytopathology Section, Laboratory of Pathology, National Cancer Institute, National Institutes of Health, Bethesda, Maryland
    • Laboratory of Pathology, National Cancer Institute, NIH, Building 10 Room 2A19, 10 Center Dr., Bethesda, MD 20892-1500
    Search for more papers by this author
    • Fax: (301) 402-2585

  • This article is a US Government work and, as such, is in the public domain in the United States of America.



Involvement of the cerebrospinal fluid (CSF) by hematopoietic malignancies may be difficult to document by morphology alone. In cases with low numbers of cells or ambiguous morphology, the diagnoses of “atypical” or “suspicious” may be used. The significance of these diagnostic terms in this scenario has not been well established.


Between January 2000 and July 2004, 32 patients with known lymphoma or leukemia and an initial diagnosis of “atypical” or “suspicious” using morphologic criteria were identified. Subsequent flow cytometry (FC) and cytologic data from these patients were evaluated.


Of the 32 patients with an initial diagnosis of “atypical” or “suspicious,” 40.6% (n = 13) had negative first and subsequent FC and morphologic evaluation of their CSF samples with follow-up up to 1 year. Nineteen patients (59.4%) had malignant hematopoietic cells identified in subsequent CSF samples by cytology and/or FC.


In patients with a previous history of lymphoma or a hematopoietic malignancy, a majority of the patients (59.4%) with an “atypical” or “suspicious” diagnosis on CSF will ultimately have malignant cells identified in the CSF by cytology and/or FC. Many of these patients can be identified more expediently with the concurrent utilization of flow cytometry. Cancer (Cancer Cytopathol) 2006. Published 2006 by the American Cancer Society.

The evaluation of cerebral spinal fluid (CSF) can be a critical test for the documentation and treatment of primary and secondary neoplastic involvement of the central nervous system (CNS). Secondary spread to the CNS of peripheral neoplasms, whether hematopoietic or solid, is a well-described phenomenon associated with poor prognosis. Lymphomas that involve the CNS tend to be high grade and include aggressive B-cell neoplasms, lymphomas associated with immunodeficiency states, acute lymphoblastic leukemia, human T-cell leukemia virus (HTLV)-1-associated T-cell neoplasia, and primary CNS lymphoma.

Historically, the monitoring of CNS involvement with any type of neoplasia has been a combination of imaging and CSF studies that include cytological and chemical evaluations. More recently, flow cytometry (FC) and molecular diagnostics are having increasing roles in detecting small clones of cells that may not be identified by cytologic evaluation alone.1–3 The cytologic identification of malignant hematopoietic cells within the CSF of patients with known CNS involvement, after 3 or more submitted samples, is reported to be approximately 93% of true-positive cases.4 In samples with low numbers of cells or with only a slight degree of cytologic atypia, it may be difficult to lend a definitive malignant or benign categorization. In circumstances such as this, a diagnosis of “atypical” or “suspicious” lymphoid cells may be rendered.

In the present study, we examined the identification of malignancy in CSF specimens from patients with primary CNS lymphoma or patients under evaluation for CNS involvement by their systemic lymphoma or leukemia. CSF specimens diagnosed as “atypical” or “suspicious” were followed for up to 1 year to determine the significance of these diagnoses as they pertain to eventual CNS/CSF involvement by lymphoma or leukemia and the ability to diagnose CNS involvement using cytological criteria and FC.


Patient Selection

Our computer files were queried for initial CSF samples in patients with known lymphomas or leukemias that were cytologically designated as “atypical” or “suspicious” between January 2000 and July 2004. Of these, 32 patients were given an initial diagnosis of “atypical” or “suspicious.” Patients with a positive CSF diagnosis (by FC or morphology) before the diagnosis of “atypical” or “suspicious” or the diagnosis of lymphomatoid granulomatosis were excluded from the study. The morphologic consideration of “atypical lymphoid cells” consisted of large lymphocytes with increased nuclear to cytoplasmic (N:C) ratios, convoluted nuclei, basophilic cytoplasm, and prominent nucleoli. In a few cases these atypical changes were thought to be of a reactive rather than a neoplastic etiology and the categorization of “atypical favor reactive” was utilized. The morphologic features of “suspicious” lymphoid cells are not well defined; however, in our laboratory and for this study, cells were categorized as “suspicious” if the cytologic features of malignancy were more distinct than in those cells deemed “atypical” (i.e., more exaggerated increased N:C ratios, nuclear membrane irregularities, and more prominent nucleoli). In most cases, the morphologic features of “suspicious” cells were similar to but less marked than the malignant cells identified in the patient's peripheral tissue specimens. Alternatively, the diagnosis of “suspicious” was used if the number of cells in question was very small. A diagnosis of “atypical” or “suspicious” was rendered in the absence of FC information; however, subsequent FC data, when available, were considered before a final diagnosis was reached. The issue of interobserver variability between pathologists was not addressed in this study. Patient outcomes were followed for up to 1 year to determine if subsequent CSF samples were eventually diagnosed as positive or negative by FC and/or cytology. Statistical analysis was performed using the InStat statistical software package (GraphPad Software, San Diego, CA). Briefly, data were entered and the mean, standard deviation, and standard error of the mean (SEM) determined.

Cytologic Analysis

Fresh CSF specimens were refrigerated and processed within several hours of collection. Each cytospin slide was prepared by centrifuging 0.5 mL of undiluted specimen at 500 revolutions per minute for 5 minutes. The slides were then air-dried and stained with Diff-Quik (Dade, Aguada, PR). Samples were concentrated or diluted in RPMI-1640 (Gibco BRL, Grand Island, NY) when required.

Each case was evaluated independently by at least 1 cytopathologist (A.A. or A.F.). To be eligible for this study, cases needed a diagnostic categorization of “atypical” or “suspicious” lymphoid cells. Immunocytochemistry was not used in this study. Our laboratory utilizes flow cytometry for the characterization of hematopoietic processes in the CSF due to the high sensitivity and specificity of the technology combined with the low number of cells necessary to detect a clone (80-100 cells to detect a clone; unpublished data).

FC Analysis

CSF samples were washed with phosphate-buffered saline to remove cytophilic antibodies before cell counting. A hemocytometer and Trypan blue exclusion were used to assess cell number and viability, respectively. Specimens were stained for 30 minutes at room temperature with an antibody cocktail containing 3-4 antibodies. Final dilution was set at the manufacturer's recommendations. Samples contaminated with red blood cells were lysed with Immunolyse (Beckman-Coulter, Miami, FL). The samples were subsequently fixed in 1.0% paraformaldehyde and stored at 4°C for up to 12 hours.

The antibody panel used was dependent on the patient's diagnosis and clinical history. The antibody panel included CD3, CD4, CD7, and CD25 for T-cells, CD56 and CD16 for natural killer cells, CD19 or CD20 for B-cells, and CD45 for leukocytes. Antibodies against the kappa and lambda light chains were paired with CD19, CD20, CD22, CD38, CD45, CD5, and CD10 based on the diagnosis of the systemic disease.

Three-color and 4-color cytometry was performed on a BD (Becton-Dickinson, San Jose, CA) FACS Caliber FC. Sensitivity of the fluorescent detectors was set and monitored by using Calibrite beads (Becton-Dickinson). Cell populations were analyzed by gating on forward scatter, side scatter, CD45, CD19, CD20, and/or CD22.


The patient population consisted of 19 males and 13 females with a mean approximate age of 45 years (range, 12-79 years). An average of 13.8 CSF samples were submitted per patient (range, 2-62 CSF samples). A summary of the diagnoses is presented in Table 1.

Table 1. Summary of Diagnoses
DiagnosesNo. of Patients
  1. CNS: central nervous system; HIV: human immunodeficiency virus; HTLV: human T-cell leukemia virus.

Acute lymphoblastic leukemia2
B-cell lymphoma1
Burkitt lymphoma2
Chronic lymphocytic leukemia2
Primary CNS lymphoma2
Diffuse large B-cell lymphoma8
Follicular lymphoma1
Gamma-delta T-cell lymphoma2
HIV non-Hodgkin lymphoma4
HTLV-1 leukemia/lymphoma2
Mycosis fungoides1
T-cell lymphoma/neoplasm5
 Total = 32 patients

The patients were categorized into 2 groups depending on the presence of malignant lymphocytes in the CSF via follow-up by FC and/or cytology (Table 2). In Group I, atypical or suspicious CSF samples from 13 (40.6%) of the 32 patients were subsequently negative for malignant cells as assessed by cytology and/or FC. The average age of these patients was 46.6 years (range, 18-79 years), with an average of 6.3 CSF samples (range, 2-19 samples) submitted per patient. The diagnosis of atypical or suspicious was made on the first submitted CSF sample in 8 of the 13 patients (61.5%). In the remaining 5 patients the diagnosis of atypical or suspicious was made after an average of 3.6 CSF specimens were submitted (range, 2-7 specimens). In these patients, an average of 52.4 days (range, 7-108 days) elapsed between the submission of the initial CSF specimen to the diagnosis of atypical or suspicious. The CSF of these patients remained free of malignant lymphocytes up to 1 year after the diagnosis of atypical or suspicious.

Table 2. Categorization of Patients by Presence of Malignant Lymphocytes and by Method of Detection
Total Sample Size n = 32GroupMean No. of CSF Specimens/PatientMalignant Cells Identified by FCMalignant Cells Identified by Cytology and FC
  1. CSF: cerebrospinal fluid; FC: flow cytometry.

Atypical/suspicious cells with negative follow-up (n = 13; 40.6%)IAverage = 6.3  
  Range: 2-19 specimens/ptNoNo
Atypical/suspicious cells with malignant cells identified by FC (n = 19; 59.4%)IIAverage = 18.8  
  Range: 2-62 specimens/ptYes (n = 19 of 19; 100%)Yes (n = 9 of 19; 47.4%)

CSF samples from patients placed in Group II were all found to be subsequently positive by FC. Morphologic evidence of malignant cells was eventually found in the CSF of 9 of the 19 (47.4%) patients included in Group II (Fig. 1). A positive diagnosis of malignancy was simultaneously made by microscopy and FC in 2 of the 9 patients. In 1 instance microscopic identification of malignant cells was made before subsequent confirmation by FC. In the remaining patients (6 of 9 patients) the diagnosis was first established by FC. For the group, an average of 79 days (range, 2-266 days) elapsed from the date of submission of the initial CSF specimen to the morphologically based diagnosis of positive for malignant cells (Table 2).

Figure 1.

Photomicrograph illustrating lymphoid cells in the cerebrospinal fluid (CSF) from a patient with systemic diffuse large B-cell lymphoma (DLBCL). (A) A lymphocyte that is typical of an “atypical” lymphocyte. Note the irregular nuclear contours and presence of nucleoli. Although abnormal, the cell itself is not large and does not fulfill the diagnostic criteria of malignancy. (B) A later sample from the same patient demonstrating a benign lymphoid cell and a large malignant cell (arrow). The large lymphoid cell exhibits a dark-blue cytoplasm, irregular nuclear contours, and cytoplasmic blebs. This cell also resembled those described in a peripheral specimen. (Diff-Quik stain ×1000).

In the CSF samples from patients confirmed by FC to contain malignant lymphocytes (Group II), an average of 31.7 days (range, 0-160 days) elapsed from the date of submission of the first CSF specimen for FC analysis to the diagnosis of positive for malignant lymphocytes; 73.7% (14 of 19 specimens) of the CSF specimens submitted for FC were diagnosed as malignant on the initial specimen. Compared with the time it took to discover malignant cells by morphology, FC took 2.49-fold less time (P≤.02) (i.e., 31.7 days vs. 79 days for FC and cytology, respectively).

Interestingly, of the patients that were positive by both cytology and FC, 7 of the 9 (77.8%) specimens involved a B-cell process and 2 of the 9 (22.2%) specimens involved a T-cell process. In comparison, of the samples that were positive by FC alone, 6 of the 10 (60%) specimens involved a B-cell process and 4 of the 10 (40%) specimens involved a T-cell process (Table 3).

Table 3. Percentage of B-cell and T-cell Lymphoproliferative Disorders in Malignant CSF
 B-Cell LymphomaT-Cell Lymphoma
  1. CSF: cerebrospinal fluid; FC: flow cytometry.

Positive cytology and FC (n = 9; 47.4%)7 of 9 (77.8%)2 of 9 (22.2%)
Positive FC only (n = 10; 52.6%)6 of 10 (60%)4 of 10 (40%)
Total = 19 (100%)13 of 19 (68.4%)6 of 19 (31.6%)

The patients in Group II were also placed into categories based on whether the collected CSF samples were morphologically diagnosed as “atypical” or “suspicious.” Of 16 cases categorized as “atypical” an average of 20.1 CSF samples were submitted per patient over a time period ranging from 3 days to 62 days. In the 3 cases originally deemed “suspicious,” an average of 12.3 CSF samples per patient were submitted for morphologic review over a time period ranging from 4 days to 27 days. In addition, the number of days between the initial morphologic diagnosis of “atypical” or “suspicious” from the date of the first submitted CSF sample was similar (62 days and 56 days for “atypical” and “suspicious,” respectively). Although suggestive, a significant difference in these data was not detected (P≥.05).

CSFs diagnosed as “atypical” eventually demonstrated morphologically malignant lymphocytes in 8 of the 16 patients (50%). In comparison, only 1 of the 3 patients with a “suspicious” CSF sample was later diagnosed as malignant using cytologic criteria alone.

Some of the patients (7 of 32; 21.9%) were further subcharacterized as “atypical, favor reactive.” This diagnosis was given to 5 patients in Group I and 2 patients in Group II. None of the patients (0 of 7) demonstrated morphological evidence of malignancy in their CSF; however, CSF samples from 2 patients in Group II were negative by cytology, but proven positive by FC. Interestingly, the CSF from these 2 patients was involved by morphologically bland gamma-delta T-cell lymphoma (Fig. 2).

Figure 2.

Photomicrograph showing lymphoid cells in the cerbrospinal fluid (CSF) of patients diagnosed with (A) a systemic B-cell process or (B) T-cell process. These cells are mildly atypical, with slight irregularities in the nuclear membrane and perinuclear clearing (hofs). These features are more typical of a reactive lymphoid process; however, the CSF represented in Panel B was later found to be involved by gamma-delta T-cell lymphoma (Diff-Quik stain, × 1000).


The implications for discovering malignant hematopoietic cells within the CSF have obvious diagnostic, therapeutic, and prognostic significance.5, 6 For example, it may be important to distinguish between patients who have CNS involvement after systemic recurrence of a peripheral lymphoma from those who have initial CNS recurrence.6 This latter group may be more susceptible to treatment, with subsequent potential cure of their CNS disease.

The current study evaluates the utility of morphology and flow cytometry for the eventual detection of hematopoietic malignant cells in CSF; however, additional ancillary techniques such as immunocytochemistry and molecular diagnostics may be of use, depending on the clinical circumstances and the morphologic differential diagnosis.1, 7 We have recently shown that molecular diagnostic studies for hematopoietic malignancies can readily be performed on scrape lysates prepared from cytology slides.1 DNA amplification from these preparations was shown to be equivalent to that of fresh material. Using these techniques, the potential identification of malignant clones of hematopoietic origin within the CSF has been greatly expanded.

Recently, Hegde et al.8 evaluated the utility of FC and cytology in assessing CSF from patients with aggressive B-cell lymphomas. Their results showed that FC was able to detect CSF involvement in 22% of 51 newly diagnosed patients (11 patients). Cytology was only able to detect malignant cells in 1 of the 11 patients. These results corroborate other studies demonstrating the ability of FC to increase detection sensitivity when used in parallel with cytology.2, 9 This is particularly true when the number of lymphocytes in a given sample is low.8

A prospective review of our high-risk patient population revealed that in CSF samples initially diagnosed as “atypical” or “suspicious,” malignant cells were subsequently found in 59.4% of the patients, as evaluated by FC or morphology. Malignant cells were discovered an average of 79 days after the submission of the first CSF sample. Of those cases identified by FC, 47% were also eventually diagnosed by cytology. Of import, there were no statistical differences between CSF specimens categorized as “atypical” or “suspicious.” This would appear to indicate that there is no real difference between the 2 diagnoses using the criteria we used to classify cells in this study. However, the sample size of the “suspicious” category was minimal (n = 3) and may not allow an accurate comparison between the 2 groups.

The incidence of malignancy in CSF samples initially diagnosed as “atypical” or “suspicious” has not been well studied. In 1 study,10 38 of 72 (53%) CSF samples suspected of lymphoma or leukemia, including patients with diseases that usually demonstrate a high level of CNS involvement (e.g., acute lymphoblastic leukemia [ALL]), were diagnosed as containing “atypical cells.” However, only 18% of these patients were eventually found to contain malignant cells using cytological criteria. FC was not used in this study.

In a more recent study, French et al.2 reported on a series of 35 patients where the cytological diagnosis of “atypical” was made on samples collected from 7 (20%) of the patients. FC done on these samples demonstrated that only 2 of these specimens (21%) were positive for malignant lymphoid cells.

Cytologic examination of CSF samples submitted for diagnosis of non-Hodgkin lymphoma has a sensitivity of 50% to 66%.11, 12 In a recent metaanalysis, Glantz et al.4 reported cytologically identified malignant cells in 74% to 100% of patients, with known CSF involvement, after the submission of 3 or more CSF samples. Our current data show that in patients diagnosed as “atypical” or “suspicious,” and with no known CNS involvement, malignant cells are cytologically identified in approximately 28% of cases. This percentage increases to 59.4% if FC is used. Of the patients with an atypical or suspicious diagnosis, an average of 13.8 CSF samples were submitted before the identification of malignant cells in the CSF.

With the proper clinical history and an adequate number of malignant cells in a given sample, many B-cell lymphomas may be identified on CSF cytology. An exception to this may be for well-differentiated lymphomas, such as chronic lymphocytic leukemia (CLL), which may be confused with normal peripheral lymphocytes; however, CLL rarely involves the CSF.13 Similarly, T-cell lymphomas may express highly variable and pleomorphic nuclei (e.g., adult T-cell leukemia/lymphoma), making identification of malignant cells more reliable.14 Conversely, some of the T-cell lymphomas (e.g., gamma-delta T-cell lymphoma) may exhibit subtle nuclear features that may be confused with a reactive process.14–16

Lymphomas of B-cell lineage were diagnosed more frequently than lymphomas of T-cell lineage in our patient population. Of the patients in Group II who had malignant cells identified by morphology and FC, 77.8% of the cases involved a B-cell process and 22.2% involved a T-cell process.

In the current study, malignancy was diagnosed by cytology approximately 79 days after the submission of the initial CSF. In comparison, malignant cells were identified in CSF samples approximately 32 days after submission of the first CSF sample to the FC laboratory. This finding was significant and suggests that the use of FC to evaluate specimens for CNS involvement may result in fewer lumbar punctures and a more expedient diagnosis.

Another finding was that the diagnosis of “atypical, favor reactive” had clinical relevance, in that the abnormal-appearing lymphocytes were not malignant in 71.4% of the cases. The 2 cases that were negative by cytology, but later diagnosed as malignant by FC, involved bland-appearing hepatosplenic (gamma-delta) T-cell lymphomas. These results reiterate the importance of submitting samples for FC, especially when a T-cell process is suspected.

The results of the current study demonstrate that the diagnoses of “atypical” or “suspicious” resulted in the eventual detection of malignant cells in 59.4% of the patients. Moreover, we found that using FC as an adjunct to cytology resulted in malignant cells being identified faster and with fewer submitted CSF samples. We conclude that the diagnoses of “atypical” and “suspicious,” when used in a judicious manner, can provide useful clinical and therapeutic information that should raise the suspicion of CNS involvement in hematopoietic neoplasia.