Flow cytometric analysis of cerebrospinal fluid identifies patients with occult central nervous system involvement in relapsed pediatric acute lymphoblastic leukemia

Central nervous system (CNS) involvement is associated with a poor prognosis and higher risk of CNS relapse in childhood acute lymphoblastic leukemia (ALL).¹ The conventional method for identifying CNS involvement is by cytospin of cerebrospinal fluid (CSF) and microscopic examination. This method is challenged by low sensitivity and specificity for rare cells and can give rise to both false-negative and -positive results.² Several studies now support that flow cytometry (FCM) is more sensitive in detecting leukemic blasts in CSF.⁷, ⁸ FCM identifies leukemic blasts based on aberrant expression of cellular markers and is highly specific in detecting malignant cells.^2,7 At initial diagnosis, CNS involvement by FCM has been shown to be an independent prognostic factor for relapse.^8-11 Reports also show that FCM can be used for monitoring CNS disease during treatment^8,10,12,13 and ongoing clinical studies are investigating whether the clearance of leukemic blasts in the CSF during treatment holds prognostic information, similarly to minimal residual disease (MRD) monitoring in the bone marrow (BM).

Several studies have reported CSF FCM results from the time of diagnosis, but only two small studies have examined CNS involvement by FCM at the time of relapse.^13,14 We here present CSF FCM results for 39 patients with relapsed ALL of any site to investigate the diagnostic value of CSF FCM at relapse.

Patients aged 0-17.9 years at initial diagnosis of ALL were enrolled in a prospective CSF FCM study at 17 hospitals in Denmark, Finland, Lithuania, Norway, and Sweden from September 2012 to December 2022. The study was conducted in accordance with the Declaration of Helsinki and approved by the national ethical committees. Written consent was obtained from legal guardians of all participants. CSF samples were collected into Transfix® tubes (Intermedico Ltd., Hellerup, Denmark) at the first lumbar puncture and until treatment day 15 after relapse. Samples were shipped to the National Danish MRD laboratory, Rigshospitalet, Copenhagen, Denmark for FCM analysis as previously described.¹⁰ The samples were analyzed using standardized FCM panels from either the NOPHO2008 protocol (BCP: CD3/CD34/ CD19/CD10/CD38/CD20/CD45; T-ALL: CD3/CD4/CD7/CD8/ CD45/CD56) or the ALLTogether1 protocol (BCP: CD81/CD3/ CD34/CD19/CD10/CD38/CD20/CD45; T-ALL: HLA-DR/CD7/ CD3/CD14/CD33/CD8/CD4/CD45/CD56+CD16). The minimal number of aberrant cells required to define CNS positivity was ten BCP or T cells with a leukemia-associated phenotype (LAIP). In difficult cases information about aberrant marker expression by BM blasts was requested to identify the blasts. Blast count in CSF (blasts/mL) was determined as blasts with a LAIP detected by FCM divided by the volume of CSF minus 200 µL (volume of TransFix per tube). Operators examining FCM or cytospin samples were blinded to the results from the other method. FCM results were blinded to the treating physician. Statistical analysis was performed with R version 4.3.0.

We collected a total of 63 CSF samples (median volume: 2.15 mL; interquartile range [IQR], 1.8-2.3; median time to FCM analysis: 1 day; IQR, 0-2 days) from 39 patients with ALL at the time of first relapse (Table 1). The cohort included patients with isolated BM relapses (N=14), CNS-involving relapses (N=19) and relapses involving other sites (N=6). Eight patients were diagnosed with T-ALL and 31 patients with BCP-ALL. Three patients were infants (<366 days of age) at initial diagnosis, and two patients were diagnosed with Philadelphia chromosome-positive (Ph⁺) ALL. At initial diagnosis, 32 patients were treated according to NOPHO2008 (EudraCT 2008-003235-20), three patients according to ALLTogether1 (EudraCT 2018-001795-38), two patients according to Interfant-06 (EudraCT 2005-004599-19) and two patients according to EsPhALL2010 (EudraCT 2004-001647-30). At relapse, 29 of 39 patients were treated according to IntReALL2010 (EudraCT 2012-000793-30/2012-000810-12). The remaining ten patients were treated with a combination of chemotherapy, irradiation, hematopoietic stem cell transplantation, Carfilzomib and immunotherapy.

For 24 (61.5%) of the 39 patients, leukemic blasts were detected by FCM in the initial CSF sample at relapse (Table 2). The median blast count for the 24 FCM positive samples was 667 blasts/mL (IQR, 32-66,341 blasts/mL). Nineteen (48.7%) relapses were classified as CNS-involving based on cytospin results, and all cytospin-positive patients were also positive by FCM. No relapses were classified as CNS-involving based on neurological symptoms and/or MRI results only. The five relapses that were CNS positive only by FCM included three morphological BM relapses, one relapse in the BM and testis, and one relapse in the BM, mediastinum, and abdomen. The median blast count for CSF samples positive by both FCM and cytospin was 20,679 blasts/mL (IOR, 112-112,714 blasts/ mL) compared to 26 blasts/mL (IQR, 13-77 blasts/mL) for CSF samples positive only by FCM (P=0.012).

At primary diagnosis, all patients except six were classified as CNS1 based on cytospin results (Table 1). For 18 patients, we had CSF FCM data at both primary diagnosis and relapse. Eight (44.4%) patients were CNS positive by FCM at primary diagnosis, but only two patients were also cytospin positive.

Table 1.Clinical characteristics of 39 included patients at diagnosis and relapse.

At relapse, 13 of 18 patients were positive by FCM. For the six patients FCM positive at both initial diagnosis and relapse, there was no significant correlation between blast levels (rS=-0.3; P=0.68) (Online Supplementary Figure S1).

We received one or more CSF samples during the first 2 weeks of treatment from 12 of 24 FCM positive patients at relapse (Figure 1). All samples were still FCM positive at day 4 (N=5/5), 80% (N=8/10) were positive at day 8, and 85.7% (N=6/7) were positive at day 15. In comparison, only 42% (N=5/12) of the patients had a follow-up sample positive by cytospin. Among the patients remaining positive by FCM at day 15 were two Ph⁺ patients and two high hyperdiploid patients. Large blast count variation was seen between patients, and fluctuation in the blast levels were observed, since we did not receive samples from all patients at all time points (day 0: 28,366 blasts/mL; IQR, 2,635-66,341; day 4: 6,900 blasts/mL; IQR: 11-11,474; day 8: 36 blasts/mL; IQR, 23-270; day 15: 84 blasts/mL; IQR, 42-91) (Figure 1). Overall, the median blast count did however decrease from day 0 to day 15.

In the present study, we showed that 13% of the included relapses were classified as CNS positive only by CSF FCM and not by cytospin. The median blast level for CSF samples with blasts on both cytospin and FCM was about 1,000-fold higher than samples only positive by FCM. The median blast level in CSF determined by FCM was 25-fold higher at relapse than at initial diagnosis in our previously published study.¹⁰ This shows that CSF FCM identifies a subset of patients with occult CNS disease at relapse, despite the high general blast load in the CNS at relapse. This is supported by a previous CSF FCM study that analyzed samples from 13 children with isolated BM relapse, and at the first lumbar puncture 54% of these patients had blasts in their CSF by FCM.¹³

For CSF samples collected during treatment from relapse patients who were FCM positive at the first lumbar puncture, 80-90% of CSF samples remained positive throughout the first 15 days of treatment. The CSF is thus not cleared as rapidly of blasts at relapse as at initial diagnosis, where less than 10% of the samples remained positive at day 15.¹⁰ This could imply that the persistent leukemic cells at relapse are a selected subpopulation, or that they reside in a protective CNS niche. Cytospin samples collected after the initial lumbar puncture were only positive in about half of the patients, which is consistent with the significant reduction in blast levels during treatment. Even though our results and a previous study¹³ demonstrate higher sensitivity of FCM in detecting blasts in CSF during relapse treatment, the prognostic impact of this remains undetermined.

The current study covers the largest published cohort of pediatric ALL patients with CSF flow data from the time of relapse. A strength of the study is that FCM was performed in a MRD flow laboratory with staff very experienced in analyzing CSF samples. The drawbacks are that the cytospin samples were not centrally reviewed, and that follow-up samples for CSF FCM only were available in 50% of the patients with FCM-positive results at relapse. In addition, the total number of patients was too small to examine the prognostic value of low-level CNS involvement detected by FCM at relapse. Larger prospective studies powered to assess the prognostic effect of CSF FCM positivity at initial relapse and during treatment are therefore needed. Follow-up samples should be collected for all FCM-positive patients, and the slow clearance of blasts in this study show that sampling should be continued beyond day 15 of treatment. To maximize sensitivity of CSF FCM, a minimum of 1-2 mL CSF should be collected directly into TransFix tubes, and FCM should be performed by national MRD laboratories to reduce time from sampling to analysis (optimally less than 48 hours). Especially the ability to monitor blasts clearance in the CSF by FCM during treatment could be a promising tool for guiding the intensity of CNS-directed therapy at relapse, enabling a reduction in neurotoxic treatment in low-risk patients and intensified treatment in high-risk patients.

Table 2.Central nervous system classification of 39 patients by flow cytometry and cytospin at first relapse.

Figure 1.Clearance of leukemic blasts in cerebrospinal fluid at relapse during therapy. The plot shows the number of blasts/ mL determined by cerebrospinal fluid (CSF) flow cytometry (FCM) for positive patients (Pt.) with available FCM data (12 of 24) from the first 15 days of therapy after relapse. Each line designates 1 patient. To include the time point, where all blasts were cleared, data points with the value zero were shown as 0.1 on the plot. The data point at day 15 not connected with a line to a day 0 sample, designates Pt. 37 with missing data on CSF sample volume on day 0.

Footnotes

• Received January 20, 2025
• Accepted July 3, 2025

Correspondence

K. SCHMIEGELOW - kjeld.schmiegelow@regionh.dk

Disclosures

No conflicts of interest to disclose.

Contributions

Funding

The study was funded by the Danish Childhood Cancer Foundation, Nordic Cancer Union, Rigshospitalet, Swedish Childhood Cancer Foundation, Danish Cancer Research Foundation, Cancer Research Foundation of Elna and Jørgen Fagerholt Pedersen, Axel Muusfeldt’s Foundation, Anders Hasselbalch’s Fight Leukemia Foundation, Otto Christensen’s Foundation, Memorial Foundation of Merchant M. Brogaard and Wife, Research Foundation of Hasselbalch and Lykkegaard Andersen, Memorial Foundation of Eva and Henry Frænkel, Dagmar Marshall’s Foundation, King Christian the Tenth Foundation, and Memorial Trust Foundation of Thora and Viggo Grove. This work is part of the Danish nation-wide research program Childhood Oncology Network Targeting Research, Organization & Life Expectancy (CONTROL) and supported by the Danish Cancer Society (R-257-A14720) and the Danish Childhood Cancer Foundation (2019-5934 and 2020-5769).

Acknowledgments

We thank all Nordic pediatric oncology centers for providing clinical data and CSF samples for flow cytometric analysis. We are grateful to the staff at the Laboratory of Pediatric Oncology and the Department of Immunology, Rigshospitalet, for their help with sample analysis and data registration.