[Skip to Content]
[Skip to Content Landing]
Figure 1.
Variation in Biomarker Levels From Baseline to Reinduction II
Variation in Biomarker Levels From Baseline to Reinduction II

Details of timing of cerebrospinal fluid sample collection relative to the treatment phases are found in eFigure 2 and eTable 3 in the Supplement. Data are expressed as means; errors bar represent 95% CIs. GFAP indicates glial fibrillary acidic protein; MBP, myelin basic protein; and NGF, nerve growth factor.

Figure 2.
Association Between Intrathecal Methotrexate and Change in Nerve Growth Factor (NGF) From Baseline to Consolidation
Association Between Intrathecal Methotrexate and Change in Nerve Growth Factor (NGF) From Baseline to Consolidation

Analyses were adjusted for age at diagnosis and baseline level of each biomarker. Intrathecal methotrexate dose is qualified in milliliters in which 1 mL contains 1 mg of methotrexate. Data points indicate unique measurements for each patient: Solid line indicates the regression line showing the association between change in NGF and intrathecal methotrexate.

Figure 3.
Association of Total Tau Level During Active Treatment Phase and Neurocognitive/Neuroimaging Outcomes at Long-term Follow-up
Association of Total Tau Level During Active Treatment Phase and Neurocognitive/Neuroimaging Outcomes at Long-term Follow-up

The change in level of total tau was measured from baseline to consolidation; long-term neurocognitive function and neuroimaging outcomes at 5 or more years after diagnosis. A, Higher age-adjusted z scores for these measures were indicative of better functioning. All models were adjusted for age at diagnosis and corrected for multiple comparisons. Details of associations between change in biomarkers from baseline to consolidation and other neurocognitive measures are described in eTable 11 in the Supplement. B, All models are adjusted for age at diagnosis and age at evaluation and corrected for multiple comparisons. Details of associations between change in biomarkers from baseline to consolidation and neuroimaging indices on the frontal and parietal lobes are described in eTable 13 in the Supplement.

Table 1.  
Change in Biomarker Levels and Risk for Leukoencephalopathya
Change in Biomarker Levels and Risk for Leukoencephalopathya
Table 2.  
Multivariable Models of Genetic Polymorphism and Change in CSF Biomarker During Treatment Phasea
Multivariable Models of Genetic Polymorphism and Change in CSF Biomarker During Treatment Phasea
1.
Krull  KR, Cheung  YT, Liu  W,  et al.  Chemotherapy pharmacodynamics and neuroimaging and neurocognitive outcomes in long-term survivors of childhood acute lymphoblastic leukemia.  J Clin Oncol. 2016;34(22):2644-2653.PubMedGoogle ScholarCrossref
2.
Cheung  YT, Krull  KR.  Neurocognitive outcomes in long-term survivors of childhood acute lymphoblastic leukemia treated on contemporary treatment protocols: a systematic review.  Neurosci Biobehav Rev. 2015;53:108-120.PubMedGoogle ScholarCrossref
3.
Skillbäck  T, Farahmand  BY, Rosén  C,  et al.  Cerebrospinal fluid tau and amyloid-β1-42 in patients with dementia.  Brain. 2015;138(pt 9):2716-2731.PubMedGoogle ScholarCrossref
4.
Rosén  C, Andersson  C-H, Andreasson  U,  et al.  Increased levels of chitotriosidase and YKL-40 in cerebrospinal fluid from patients with Alzheimer’s disease.  Dement Geriatr Cogn Dis Extra. 2014;4(2):297-304.PubMedGoogle ScholarCrossref
5.
Wang  J, Zou  Q, Han  R, Li  Y, Wang  Y.  Serum levels of glial fibrillary acidic protein in Chinese children with autism spectrum disorders.  Int J Dev Neurosci. 2017;57:41-45.PubMedGoogle ScholarCrossref
6.
Crawford  JD, Chandley  MJ, Szebeni  K, Szebeni  A, Waters  B, Ordway  GA.  Elevated GFAP protein in anterior cingulate cortical white matter in males with autism spectrum disorder.  Autism Res. 2015;8(6):649-657.PubMedGoogle ScholarCrossref
7.
Krawczuk-Rybak  M, Grabowska  A, Protas  PT, Muszynska-Roslan  K, Holownia  A, Braszko  J.  Intellectual functioning of childhood leukemia survivors: relation to tau protein—a marker of white matter injury.  Adv Med Sci. 2012;57(2):266-272.PubMedGoogle ScholarCrossref
8.
Cole  PD, Beckwith  KA, Vijayanathan  V, Roychowdhury  S, Smith  AK, Kamen  BA.  Folate homeostasis in cerebrospinal fluid during therapy for acute lymphoblastic leukemia.  Pediatr Neurol. 2009;40(1):34-41.PubMedGoogle ScholarCrossref
9.
Caron  JE, Krull  KR, Hockenberry  M, Jain  N, Kaemingk  K, Moore  IM.  Oxidative stress and executive function in children receiving chemotherapy for acute lymphoblastic leukemia.  Pediatr Blood Cancer. 2009;53(4):551-556.PubMedGoogle ScholarCrossref
10.
Krull  KR, Bhojwani  D, Conklin  HM,  et al.  Genetic mediators of neurocognitive outcomes in survivors of childhood acute lymphoblastic leukemia.  J Clin Oncol. 2013;31(17):2182-2188.PubMedGoogle ScholarCrossref
11.
Pui  CH, Campana  D, Pei  D,  et al.  Treating childhood acute lymphoblastic leukemia without cranial irradiation.  N Engl J Med. 2009;360(26):2730-2741.PubMedGoogle ScholarCrossref
12.
Pui  C-H.  Toward optimal central nervous system-directed treatment in childhood acute lymphoblastic leukemia.  J Clin Oncol. 2003;21(2):179-181.PubMedGoogle ScholarCrossref
13.
Bhojwani  D, Sabin  ND, Pei  D,  et al.  Methotrexate-induced neurotoxicity and leukoencephalopathy in childhood acute lymphoblastic leukemia.  J Clin Oncol. 2014;32(9):949-959.PubMedGoogle ScholarCrossref
14.
Kishi  S, Griener  J, Cheng  C,  et al.  Homocysteine, pharmacogenetics, and neurotoxicity in children with leukemia.  J Clin Oncol. 2003;21(16):3084-3091.PubMedGoogle ScholarCrossref
15.
National Institutes of Health; National Cancer Institute. Common Terminology Criteria for Adverse Events (CTCAE), Version 4.03. https://evs.nci.nih.gov/ftp1/CTCAE/CTCAE_4.03_2010-06-14_QuickReference_8.5x11.pdf. June 14, 2010. Accessed August 1, 2016.
16.
Benjamini  Y, Hochberg  Y.  Controlling the false discovery rate: a practical and powerful approach to multiple testing.  J R Stat Soc B. 1995;57(1):289-300.Google Scholar
17.
Rodriguez  JI, Kern  JK.  Evidence of microglial activation in autism and its possible role in brain underconnectivity.  Neuron Glia Biol. 2011;7(2-4):205-213.PubMedGoogle ScholarCrossref
18.
Bond  J, Hough  R, Moppett  J, Vora  A, Mitchell  C, Goulden  N.  ‘Stroke-like syndrome’ caused by intrathecal methotrexate in patients treated during the UKALL 2003 trial.  Leukemia. 2013;27(4):954-956.PubMedGoogle ScholarCrossref
19.
Villoslada  P, Genain  CP.  Role of nerve growth factor and other trophic factors in brain inflammation.  Prog Brain Res. 2004;146:403-414.PubMedGoogle ScholarCrossref
20.
Jiang  R, Brummett  BH, Babyak  MA, Siegler  IC, Williams  RB.  Brain-derived neurotrophic factor (BDNF) Val66Met and adulthood chronic stress interact to affect depressive symptoms.  J Psychiatr Res. 2013;47(2):233-239.PubMedGoogle ScholarCrossref
21.
Grabe  HJ, Schwahn  C, Mahler  J,  et al.  Genetic epistasis between the brain-derived neurotrophic factor Val66Met polymorphism and the 5-HTT promoter polymorphism moderates the susceptibility to depressive disorders after childhood abuse.  Prog Neuropsychopharmacol Biol Psychiatry. 2012;36(2):264-270.PubMedGoogle ScholarCrossref
22.
Yang  M, Moon  C.  Neurotoxicity of cancer chemotherapy.  Neural Regen Res. 2013;8(17):1606-1614.PubMedGoogle Scholar
23.
Wilson  CJ, Finch  CE, Cohen  HJ.  Cytokines and cognition–the case for a head-to-toe inflammatory paradigm.  J Am Geriatr Soc. 2002;50(12):2041-2056.PubMedGoogle ScholarCrossref
24.
Abbott  NJ, Rönnbäck  L, Hansson  E.  Astrocyte-endothelial interactions at the blood-brain barrier.  Nat Rev Neurosci. 2006;7(1):41-53.PubMedGoogle ScholarCrossref
25.
Wardill  HR, Mander  KA, Van Sebille  YZA,  et al.  Cytokine-mediated blood brain barrier disruption as a conduit for cancer/chemotherapy-associated neurotoxicity and cognitive dysfunction.  Int J Cancer. 2016;139(12):2635-2645.PubMedGoogle ScholarCrossref
26.
Savitz  J, Solms  M, Ramesar  R.  The molecular genetics of cognition: dopamine, COMT and BDNF.  Genes Brain Behav. 2006;5(4):311-328.PubMedGoogle ScholarCrossref
Original Investigation
July 12, 2018

Association of Cerebrospinal Fluid Biomarkers of Central Nervous System Injury With Neurocognitive and Brain Imaging Outcomes in Children Receiving Chemotherapy for Acute Lymphoblastic Leukemia

Author Affiliations
  • 1Department of Epidemiology and Cancer Control, St Jude Children’s Research Hospital, Memphis, Tennessee
  • 2Department of Neurology, St Jude Children’s Research Hospital, Memphis, Tennessee
  • 3Department of Biostatistics, St Jude Children’s Research Hospital, Memphis, Tennessee
  • 4Department of Psychology, St Jude Children’s Research Hospital, Memphis, Tennessee
  • 5Center for Personalized Medicine, University of Colorado Anschutz Medical Campus, Denver
  • 6Department of Diagnosis Imaging, St Jude Children’s Research Hospital, Memphis, Tennessee
  • 7Pediatric Blood and Bone Marrow Transplant Program, University of Minnesota, Minneapolis
  • 8Division of Pulmonary, Allergy, Critical Care and Sleep Medicine, University of Minnesota, Minneapolis
  • 9Department of Oncology, St Jude Children’s Research Hospital, Memphis, Tennessee
JAMA Oncol. 2018;4(7):e180089. doi:10.1001/jamaoncol.2018.0089
Key Points

Question  Are cerebrospinal fluid biomarkers of brain injury associated with treatment-related neurotoxic effects in children who receive chemotherapy for acute lymphoblastic leukemia?

Findings  In this cohort study of 235 children with acute lymphoblastic leukemia, methotrexate exposure was positively correlated with markers of neuronal damage, and markers of demyelination and axonal damage were associated with as much as a 70% higher risk of developing leukoencephalopathy during therapy, worse long-term frontal lobe white matter integrity, and poorer neurocognitive function. Carriers of the Val allele in the COMT gene demonstrated greater biomarker level elevations after methotrexate exposure than did carriers of the Met allele.

Meaning  Monitoring cerebrospinal fluid biomarkers and screening for genetic mediators of brain injury may help in identifying survivors at risk for abnormal neurodevelopment.

Abstract

Importance  Little is known about treatment-related neurotoxic mechanisms in children with acute lymphoblastic leukemia (ALL) treated with chemotherapy only.

Objective  To examine concentration of cerebrospinal fluid (CSF) biomarkers of brain injury at ALL diagnosis and during cancer therapy and to evaluate associations with long-term neurocognitive and neuroimaging outcomes and relevant genetic polymorphisms.

Design, Setting, and Participants  This prospective cohort study included 235 patients with ALL who received a chemotherapy-only protocol. Patients provided CSF samples after diagnosis and throughout treatment. At 5 or more years after the diagnosis, 138 (69.7%) of 198 eligible survivors participated in long-term follow-up assessments. Children were treated from June 1, 2000, through October 31, 2010. Follow-up was completed on October 21, 2014, and data were analyzed from August 1, 2015, through September 30, 2016.

Exposures  Plasma concentration of high-dose intravenous methotrexate sodium and number of triple intrathecal chemotherapy injections.

Main Outcomes and Measures  The CSF samples were assayed at 5 points from diagnosis to reinduction for biomarkers of myelin degradation (myelin basic protein [MBP]), neuronal damage (nerve growth factor [NGF] and total and phosphorylated tau protein), astrogliosis (glial fibrillary acidic protein [GFAP]), and neuroinflammation (chitotriosidase). DNA was genotyped for polymorphisms in drug metabolism, oxidative stress, and neurodevelopment. Leukoencephalopathy was evaluated by brain imaging. At 5 or more years after the diagnosis, survivors completed neurocognitive testing and brain imaging of white matter integrity.

Results  Among the 235 patients with CSF samples (120 boys [51.1%] and 115 girls [48.9%]; mean [SD] age at diagnosis, 6.8 [4.7] years), MBP and GFAP levels were elevated at baseline and through consolidation. The number of intrathecal injections was positively correlated with NGF level increase at consolidation (r = 0.19; P = .005). Increases in GFAP (risk ratio [RR], 1.23; 95% CI, 1.09-1.40), MBP (RR, 1.06; 95% CI, 1.01-1.11), and total tau (RR, 1.76; 95% CI, 1.11-2.78) levels were associated with a higher risk for leukoencephalopathy and higher apparent diffusion coefficient in frontal lobe white matter 5 years after diagnosis (standardized estimate, 0.05; P < .001). Increase in total tau at consolidation was associated with worse attention (omissions z score estimate, −0.20; P = .04).

Conclusions and Relevance  Glial injury may be present at diagnosis of ALL. Neuronal injury was associated with intrathecal chemotherapy. The CSF biomarkers may be useful in identifying individuals at risk for worse neurologic outcomes, particularly those with genetic susceptibility to poor brain function.

Introduction

Neurocognitive deficits are a common late effect found among long-term survivors of childhood acute lymphoblastic leukemia (ALL). Survivors of childhood ALL who receive chemotherapy without prophylactic cranial irradiation therapy are at risk for impairment in executive function and processing speed.1,2 Factors that increase this risk include being younger at treatment and having higher levels of exposure to high-dose intravenous methotrexate sodium and intrathecal methotrexate.1,2 Little is known about the specific sequence of cellular injury after chemotherapy, and accurate early identification of individual patients at greatest risk is still beyond current capabilities.

Biomarkers in the cerebrospinal fluid (CSF) provide information about specific cellular integrity in the central nervous system (CNS). An increase in extracellular total tau (t-tau) levels is suggestive of neuronal cell body and axonal damage, a process that underlies cognitive impairment in dementia.3 Increased chitotriosidase level in the CSF is suggestive of inflammation and microglial activation in dementia.4 Recent studies5,6 have observed elevations in glial fibrillary acidic protein (GFAP) level in autism, suggestive of astrogliosis.

Preliminary studies of CSF biomarkers7-9 suggest that changes during cancer therapy correspond to future CNS outcomes. In 31 survivors of childhood ALL who received only chemotherapy,7 increased CSF concentration of t-tau protein level after induction therapy was associated with lower intelligence 6 years after the diagnosis. Higher peak levels of t-tau during therapy were observed in 9 survivors of childhood ALL who developed leukoencephalopathy compared with 27 survivors with normal brain imaging findings.8 Increased oxidized phosphatidylcholine levels within the CSF after induction and consolidation therapy were associated with worse executive function at follow-up.9 Although the strength of this preliminary evidence is limited by small sample sizes, these findings suggest that chemotherapy affects CNS cell integrity.

Neurocognitive outcomes in survivors of childhood ALL may also be mediated by polymorphisms in genes related to antifolate chemotherapy, oxidative stress, and CNS susceptibility.10 Although polymorphisms in these genes have been examined for their contribution to posttreatment cognitive function, their association with indices of CNS integrity has not been reported. Interaction between genetic predisposition and treatment intensity may determine the severity of brain injury in ALL survivors.

We identified biomarkers related to myelin integrity, neuronal injury, astrogliosis, and neuroinflammation a priori and examined concentration of these biomarkers at diagnosis and during therapy. Based on previous findings,1 we expected these biomarker levels to increase during consolidation therapy. Associations were examined with acute leukoencephalopathy, long-term neurocognitive and neuroimaging outcomes, and polymorphisms in enzymes related to chemotherapy, oxidative stress, and CNS susceptibility.

Methods

From June 1, 2000, through October 31, 2010, 408 children with newly diagnosed ALL were treated at St Jude Children’s Research Hospital, Memphis, Tennessee, with a protocol that eliminated prophylactic cranial irradiation therapy.11 To be eligible for long-term follow-up, survivors had to be at least 5 years after diagnosis and 8 years of age. Survivors were excluded if they had a relapse and received cranial irradiation therapy or hematopoietic cell transplant, were not proficient in English, or had genetic disorders or treatment-unrelated neurologic injury associated with cognitive impairment. Among the 235 patients who provided CSF samples during active therapy, 198 met eligibility criteria for long-term follow-up and 138 (69.7%) participated (eFigure 1 in the Supplement). Demographic and clinical characteristics of the long-term survivors have been previously described as part of larger samples.1,10 This study was approved by the institutional review board at St Jude Children’s Research Hospital; written informed consent or assent was obtained from parents and participants as appropriate.

All children, regardless of risk stratification, received triple intrathecal treatments (methotrexate, hydrocortisone sodium succinate, and cytarabine), 4 courses of intravenous high-dose methotrexate chemotherapy with leucovorin calcium rescue therapy, and oral dexamethasone sodium phosphate pulses in addition to other chemotherapeutic agents (eFigure 2 in the Supplement).11,12 Leucovorin calcium rescue therapy (10-15 mg/m2 intravenously) was administered 42 hours after high-dose methotrexate infusion and repeated every 6 hours for a total of 5 doses. Exposures to intravenous high-dose methotrexate were measured as previously described.13,14 Blood samples were obtained before and at 6, 23, and 42 hours after the start of high-dose methotrexate infusion. Exposure to high-dose methotrexate was quantified as the area under the curve. Baseline, clinical, and treatment characteristics of all participants are presented in eTable 1 in the Supplement. Baseline demographics were similar between patients who provided CSF samples and survivors who participated in long-term follow-up assessments (eTable 2 in the Supplement).

Procedures

Samples of CSF were collected through spinal taps at diagnosis and before intrathecal chemotherapy administration. After lumbar puncture, CSF samples were centrifuged at 800g for 5 minutes at 4°C to remove cellular debris, divided into aliquots, and immediately frozen at −80°C. Any sample contaminated with blood was excluded from analysis. Assays were conducted on 1-mL samples collected at baseline, after induction, during consolidation, at the start of reinduction I, and during reinduction II. The study schema and points of CSF collection are summarized in eFigure 2 and eTable 3 in the Supplement, respectively.

The CSF markers of interest included levels of t-tau, phosphorylated tau (p-tau), neuron-specific enolase, nerve growth factor (NGF), GFAP, myelin basic protein (MBP), and chitotriosidase. The cellular localization and function of these markers are presented in eFigure 3 in the Supplement. All assays were conducted in duplicate at a Clinical Laboratory Improvement Amendment–approved cytokine reference laboratory using standardized immunoassays (eMethods in the Supplement).

At follow-up, survivors completed neurocognitive testing with certified examiners under the supervision of a board-certified clinical neuropsychologist (K.R.K). Testing procedures followed standard clinical guidelines. Measures of executive function, processing speed, intelligence, attention, and memory span were examined (eTable 4 in the Supplement).

A total of 182 of the 235 survivors (77.4%) who provided CSF samples completed brain magnetic resonance imaging during active therapy at as many as 4 points (eFigure 2 in the Supplement), including after induction (days 33-46), after consolidation (week 1 of reinduction), at continuation (week 48), and at end of therapy (week 120).13 The magnetic resonance images were reviewed by a board-certified neuroradiologist, who graded leukoencephalopathy according to version 4.03 of the Common Terminology Criteria for Adverse Events (CTCAE).15 Axial T2-weighted and axial T2 fluid-attenuated inversion recovery images were used for grading. The neuroradiologist was blinded to levels of CSF biomarkers and neurocognitive outcomes.

Diffusion tensor imaging was performed using a 1.5-T high-resolution 3-dimensional T1-weighted scan (approximately 1 mm3) using a double spin-echo echoplanar imaging pulse sequence (repetition time, 10 000 milliseconds; echo time, 100 milliseconds; matrix, 3 × 1.8 × 1.8 mm) with 4 acquisitions and 12 gradient directions. Diffusion tensor imaging was performed during the same visit as the neurocognitive testing in 105 of the 138 survivors (76.1%) who participated in long-term follow-up at more than 5 years after diagnosis. Measures of fractional anisotropy and apparent diffusion coefficient were obtained from voxels in frontal and parietal lobes. These regions were selected a priori; activity in these regions was expected to correspond with neurocognitive measures.1

Genotyping was conducted in a previous study that reported associations with end-of-therapy attention problems.10 Forty-two single-nucleotide polymorphisms (SNPs) were selected a priori based on literature linking them to methotrexate metabolism and folate pathway, general drug metabolism, oxidative stress, and attention or executive function problems in the general pediatric population.10 Frequency results for the 42 SNPs are listed in eTable 5 in the Supplement.

Statistical Analysis

Follow-up was completed on October 21, 2014, and data were analyzed from August 1, 2015, through September 30, 2016. We used descriptive analysis to summarize CSF biomarker levels during the course of therapy. For all analyses, biomarkers were treated as continuous variables. Given the pattern of change observed across time, we focused on the change from baseline to consolidation, which corresponds to after administration of the first intrathecal therapy and the first cycle of high-dose intravenous methotrexate. We used general linear modeling to evaluate associations of high-dose intravenous methotrexate and intrathecal therapy with change in biomarker levels from baseline to consolidation, adjusting for age at diagnosis and baseline level of the biomarker.

We conducted a Poisson regression model to estimate risk ratios and 95% CIs for development of postinduction leukoencephalopathy in response to changes in levels of CSF biomarkers from baseline to after induction and postconsolidation leukoencephalopathy from changes in levels from after induction to consolidation. Leukoencephalopathy was coded as present or absent, given that most CTCAE grades were 1. Models were adjusted for age at diagnosis and baseline CSF levels.

Long-term neurocognitive outcomes were transformed into age-adjusted z scores based on nationally representative normative data. Only neurocognitive measures with scores significantly lower than expected (mean, 0; SD, 1) on a 1-sample t test at P < .05 after correcting for false discovery rate were included in subsequent analyses. We conducted general linear modeling to evaluate the association between CSF marker levels at baseline and change from baseline to consolidation with neurocognitive outcomes during long-term follow-up, adjusting for the false discovery rate. Models were adjusted for age at diagnosis.

A similar approach was adopted with long-term neuroimaging measures, with fractional anisotropy and apparent diffusion coefficient as outcomes adjusted for age at diagnosis and age at evaluation. Associations were examined among each SNP and in change in CSF biomarker levels from baseline to consolidation using an additive model adjusted for baseline levels. Single-nucleotide polymorphisms that were significantly associated with change at the level of P < .10 in univariate analyses were included in multiple general linear models. Multivariable models were adjusted for age at diagnosis, race, and baseline level of the CSF biomarker.

The interaction between high-dose intravenous methotrexate and SNPs on change in CSF levels was also tested. A correlation coefficient was presented to denote the strength of the association between CSF markers and outcomes. All analyses were conducted using SAS software (version 9.3; SAS Institute Inc). Homogeneity of variance and normality of residuals were checked to ensure that model assumptions were met. All P values from analyses of change in biomarker levels over time and biomarker associations with survivor characteristics, treatment, neurocognitive outcomes, and brain imaging outcomes were adjusted for false discovery rate using MULTTEST in SAS software and a linear step-up method.16 Voxelwise tensor calculations in magnetic resonance imaging were performed with SPM8 software (http://www.fil.ion.ucl.ac.uk/spm/software/spm8/). A summary of all statistical analyses is found in eTable 6 in the Supplement.

Results

Among the 235 patients with CSF samples available (120 boys [51.1%] and 115 girls [48.9%]; mean [SD] age at diagnosis, 6.8 [4.7] years), substantial changes and interpatient variability in levels of CSF biomarkers were found during the course of treatment. Concentrations of t-tau (mean change, 0.13 ng/mL; 95% CI, 0.04-0.22 ng/mL; P < .001), p-tau (mean change, 0.36 nM; 95% CI, 0.25-0.47 nM; P < .001), neuron-specific enolase (mean change, 13.2 ng/mL; 95% CI, 10.8-15.6 ng/mL; P < .001), NGF (mean change, 1.71 ng/mL; 95% CI, 0.30-3.12 ng/mL; P < .001), and chitotriosidase (mean change, 354.02 mmol/mL/h; 95% CI, 60.23-547.80 mmol/mL/h; P < .001) significantly increased from baseline to consolidation (Figure 1). The GFAP level fluctuated throughout treatment, but GFAP levels at baseline and reinduction II remained similar. The MBP levels were elevated at baseline and decreased from consolidation to reinduction II. Most of the biomarker levels returned to undetectable or baseline levels at reinduction I with the exception of GFAP. Concentrations of t-tau, chitotriosidase, and GFAP remained detectable in most of the survivors at reinduction II (eTable 7 in the Supplement). Before initiation of treatment, older age at diagnosis was associated with higher concentrations of neuron-specific enolase (r = 0.25; P < .001), and female survivors had higher concentrations of NGF (mean [SD], 0.08 [0.04] vs 0.04 [0.04] ng/mL; P = .04) (eTable 7 in the Supplement). Risk stratification was not associated with baseline biomarker levels, but higher baseline chitotriosidase level was associated with adverse cytogenetic factors (mean [SD], 90.4 [197.3] vs 20.9 [53.9] mmol/mL/h; P = .002). Survivors with TELAML1 fusion gene (OMIM 151385) mutations had lower levels of neuron-specific enolase (mean [SD], 18.8 [12.5] vs 23.2 [13.6] ng/mL; P = .048) and MBP (mean [SD], 9.1 [7.1] vs 12.8 [8.6] ng/mL; P = .01) (eTable 8 in the Supplement).

During consolidation, the dose of intrathecal methotrexate was associated with increased NGF levels (standardized estimate, 2.83; P = .005) (Figure 2), although the association was of modest strength (r = 0.19). Greater exposure to high-dose methotrexate was marginally associated with increased GFAP levels (standardized estimate, 1.85; P = .07). We found no other associations between treatment exposures and biomarkers (eTable 9 in the Supplement).

Incidence of leukoencephalopathy (42 patients [23.3%]) during the course of treatment was reported previously.13 Most cases of detected leukoencephalopathy were asymptomatic. From baseline to after induction, increased concentrations of GFAP were associated with a higher risk ratio (RR) (1.23; 95% CI, 1.09-1.40) of leukoencephalopathy after induction (Table 1). From after induction to consolidation, increases in concentrations of t-tau (RR, 1.76; 95% CI, 1.11-2.78) and MBP (RR, 1.06; 95% CI, 1.01-1.11) were associated with a higher risk of developing leukoencephalopathy at the end of the consolidation.

At long-term follow-up, survivors’ neurocognitive performance decreased below the population mean on multiple measures (eTable 10 in the Supplement). Greater increase in concentrations of t-tau from baseline to consolidation was associated with poorer performance on measures of attention at follow-up (omissions z score estimate, −0.20; P = .04) (Figure 3). Few associations between other biomarkers and neurocognitive outcomes were identified (eTables 11 and 12 in the Supplement).

Greater change in t-tau level from baseline to consolidation was associated with worse white matter integrity in frontal lobes during long-term follow-up, as reflected by higher apparent diffusion coefficient (standardized estimate, 0.05; P < .001) (Figure 3). Associations were found between changes in the frontal and parietal lobes in NGF (frontal estimate, 0.003 [P = .15]; parietal estimate, 0.003 [P = .04]) and MBP (frontal estimate, 0.002 [P = .04]; parietal estimate, 0.002 [P = .04]) levels and white matter indices (eTables 13 and 14 in the Supplement).

Associations were examined between polymorphisms and CSF biomarkers related to treatment exposures (GFAP and NGF) or long-term neurologic outcomes (t-tau, neuron-specific enolase, and MBP). After adjustment for age at diagnosis, race, and baseline concentration of CSF biomarker, change in GFAP levels from baseline to consolidation was associated with polymorphisms in genes related to CNS susceptibility, including APOE (OMIM 107741) Cys112Arg (standardized estimate, 0.27; P = .04) (Table 2). Significant associations between Val-Met variations in the COMT gene and decreased concentrations of GFAP were also observed (standardized estimate, −0.18; P = .04) (eFigure 4 in the Supplement).

Discussion

This study provides novel insight into CNS injury during chemotherapy for childhood ALL. A biomarker suggestive of breakdown of the astrocyte cytoskeleton (GFAP) was associated with exposure to higher doses of intravenous methotrexate, whereas a biomarker suggestive of neuronal cell injury (NGF) was related to intrathecal methotrexate therapy. Breakdown in the integrity of axons, reflected through increase in MBP levels, was associated with subsequent development of acute leukoencephalopathy and myelin integrity at long-term follow-up. Increase in t-tau levels, reflective of axonal injury and plasticity in developmental models, was predictive of future leukoencephalopathy and poorer attention at long-term follow-up. This pattern suggests injury to brain glia and neurons early during chemotherapy, with effects lasting into long-term follow-up.

Most biomarker levels decreased to undetectable by the end of therapy with the exception of t-tau, GFAP, and chitotriosidase. This finding may indicate persistent astrogliosis and microglial activation in response to CNS inflammation. Previous studies17 have highlighted early involvement of microglial and immune-related changes in neurodegenerative diseases such as Alzheimer disease. Our results suggest that neuroinflammation may play a role in the neurotoxic sequelae of chemotherapy. Persistent astrogliosis and microglial activation may contribute to persistent CNS pathologic features. This proposed mechanism of CNS injury should be further validated.

Consistent with literature on cognitive impairment,7 t-tau is associated with leukoencephalopathy during therapy and white matter microstructure at long-term follow-up. Higher levels of CSF tau protein are an indicator of axonal injury and cell death. We observed associations between increased t-tau level and problems with attention, processing speed, and white matter integrity in frontal lobes at long-term follow-up. Our results suggest that axons are injured early during chemotherapy, and this injury continues to be reflected in measures of brain integrity at long-term follow-up. Monitoring t-tau level as part of clinical practice may help identify subgroups of patients at risk for adverse neurologic outcomes.

Intrathecal methotrexate therapy was associated with an increase in NGF level. Methotrexate administered with cytarabine increases sensitivity to glutamate-mediated neuronal damage.18 Increased NGF level after administration of intrathecal methotrexate may suggest neuronal injury in response to acute inflammation.19 Increased NGF level during chemotherapy was associated with poorer white matter integrity in frontal and parietal lobes during long-term follow-up, likely a function of lower myelin development in injured neural pathways. Survivors with the Val66Met mutation on the BDNF gene (OMIM 113505) had a marginally higher elevation in NGF levels compared with Met carriers, consistent with evidence suggesting Met-associated protective mechanisms in populations without cancer.20,21

The importance of glial cells, such as oligodendrocytes, astrocytes, and microglia, in chemotherapy-induced neurotoxicity has recently surfaced.22 Studies in elderly persons suggest that inflammation within the CNS contributes to cognitive impairments through interactions between neurons and glial cells.23 Our results suggest a dose-dependent increase in GFAP levels after administration of high-dose methotrexate. As the providers of energy substrates, astrocytes establish a direct link between the blood-brain barrier and neurons.24 Proinflammatory cytokines, which can be induced by chemotherapy, have detrimental effects on tight junctions and thus the integrity of the blood-brain barrier.25 Circulating cytokines may cross the blood-brain barrier and elicit a cascade of neurotoxic events in the CNS.

We observed large variations in concentrations of CSF markers at baseline, which may suggest the presence of CNS injury before chemotherapy. High levels of baseline MBP and GFAP may suggest that demyelination and glial cell injury have occurred owing to acute inflammatory responses induced by the cancer. Future studies should evaluate the trajectory and association among neuroinflammation, oxidative stress, and neuron injury throughout treatment. Chitotriosidase, a marker of microglial activation, was associated with genetic prognostic markers of childhood ALL. Because the initial risk status and disease presentation were not significantly associated with baseline levels, the interpatient variation may be attributed to genetic susceptibility. Carriers of the Val allele in the COMT gene demonstrated more CNS injury after high-dose methotrexate treatment. These carriers have greater degradation of dopamine,26 especially in the prefrontal cortex. Genetic polymorphisms that regulate dopamine and synaptosome-associated protein were associated with CSF biomarkers. Larger elevation in GFAP concentrations was found in survivors with the risk allele in the APOE gene, which is associated with attention problems in survivors of childhood ALL.10 The roles of COMT and APOE genes in CNS injury and functional outcomes and interactions with treatment and environmental factors need further validation within this population.

Limitations

Several limitations should be considered in interpreting these results. Determining differential effects and interactions between chemotherapeutic agents is difficult. Serum assays for pharmacokinetics and biomarkers were conducted at specific cycles or a single point. We could not determine the onset of leukoencephalopathy during the therapy because subclinical white matter changes may have occurred before the brain magnetic resonance imaging was conducted. Still, the variables should reflect prototypical responses to treatment during the course of therapy. Interpretation of biomarker level is limited by the lack of noncancer controls. Given that the current sample size of survivors is larger than those of most existing studies, we could compare outcomes across survivors with varying intensities of treatment exposures. Although modest associations between treatment exposures and biomarkers were identified, effect sizes (eTable 8 in the Supplement) are substantial for certain biomarkers. Given relatively rare minor allele frequency for some polymorphisms, we were not able to examine the difference between heterozygous and homozygous effects in gene-outcome associations. This overall pattern of findings requires validation in larger prospective studies, including examination of recessive gene variants.

Conclusions

Our results suggest that CSF biomarkers of CNS injury have clinically relevant associations with treatment characteristics and long-term outcomes. Monitoring levels of t-tau within CSF may estimate the risk of leukoencephalopathy and abnormal neurodevelopment. Coupled with knowledge of genetic mediators of CNS injury, the present results may advance the development of personalized medicine whereby treatments for and prevention of late effects are tailored to characteristics of individuals.

Back to top
Article Information

Accepted for Publication: January 5, 2018.

Corresponding Author: Kevin R. Krull, PhD, Department of Epidemiology and Cancer Control, St Jude Children’s Research Hospital, 262 Danny Thomas Pl, MS 735, Memphis, TN 38105 (kevin.krull@stjude.org).

Published Online: March 29, 2018. doi:10.1001/jamaoncol.2018.0089

Author Contributions: Dr Krull had full access to all the data in the study and takes responsibility for the integrity of the data and the accuracy of the data analysis.

Study concept and design: Cheung, Khan, Edelmann, Hudson, Krull.

Acquisition, analysis, or interpretation of data: All authors.

Drafting of the manuscript: Cheung, Edelmann, Srivastava, Cheng, Krull.

Critical revision of the manuscript for important intellectual content: Cheung, Khan, Liu, Brinkman, Edelmann, Reddick, Pei, Panoskaltsis-Mortari, Cheng, Robison, Hudson, Pui, Krull.

Statistical analysis: Cheung, Liu, Edelmann, Pei, Srivastava, Cheng, Krull.

Obtained funding: Hudson, Krull.

Administrative, technical, or material support: Reddick, Panoskaltsis-Mortari, Robison, Pui, Krull.

Study supervision: Cheung, Krull.

Conflict of Interest Disclosures: None reported.

Funding/Support: This study was supported by grant MH085849 from the National Institute of Mental Health (Dr Krull), grant CA195547 from the National Cancer Institute (Drs Robison and Hudson), and the American Lebanese Syrian Associated Charities.

Role of the Funder/Sponsor: The sponsors had no role in the design and conduct of the study; collection, management, analysis, and interpretation of the data; preparation, review, or approval of the manuscript; and decision to submit the manuscript for publication.

Additional Contributions: Mary V. Relling, PharmD, St Jude Children’s Research Hospital, and her laboratory provided the pharmacokinetic data. Cara Kimberg, PhD, Cynthia Jones, MA, Deborah Stewart, MA, and Adrienne Studaway, MEd, St Jude Children’s Research Hospital, administered the neurocognitive tests. Joycelynn Butler, MS, and Eric Caron, APN, MSN, St Jude Children’s Research Hospital, extracted the data. None of these contributors were compensated for this work.

References
1.
Krull  KR, Cheung  YT, Liu  W,  et al.  Chemotherapy pharmacodynamics and neuroimaging and neurocognitive outcomes in long-term survivors of childhood acute lymphoblastic leukemia.  J Clin Oncol. 2016;34(22):2644-2653.PubMedGoogle ScholarCrossref
2.
Cheung  YT, Krull  KR.  Neurocognitive outcomes in long-term survivors of childhood acute lymphoblastic leukemia treated on contemporary treatment protocols: a systematic review.  Neurosci Biobehav Rev. 2015;53:108-120.PubMedGoogle ScholarCrossref
3.
Skillbäck  T, Farahmand  BY, Rosén  C,  et al.  Cerebrospinal fluid tau and amyloid-β1-42 in patients with dementia.  Brain. 2015;138(pt 9):2716-2731.PubMedGoogle ScholarCrossref
4.
Rosén  C, Andersson  C-H, Andreasson  U,  et al.  Increased levels of chitotriosidase and YKL-40 in cerebrospinal fluid from patients with Alzheimer’s disease.  Dement Geriatr Cogn Dis Extra. 2014;4(2):297-304.PubMedGoogle ScholarCrossref
5.
Wang  J, Zou  Q, Han  R, Li  Y, Wang  Y.  Serum levels of glial fibrillary acidic protein in Chinese children with autism spectrum disorders.  Int J Dev Neurosci. 2017;57:41-45.PubMedGoogle ScholarCrossref
6.
Crawford  JD, Chandley  MJ, Szebeni  K, Szebeni  A, Waters  B, Ordway  GA.  Elevated GFAP protein in anterior cingulate cortical white matter in males with autism spectrum disorder.  Autism Res. 2015;8(6):649-657.PubMedGoogle ScholarCrossref
7.
Krawczuk-Rybak  M, Grabowska  A, Protas  PT, Muszynska-Roslan  K, Holownia  A, Braszko  J.  Intellectual functioning of childhood leukemia survivors: relation to tau protein—a marker of white matter injury.  Adv Med Sci. 2012;57(2):266-272.PubMedGoogle ScholarCrossref
8.
Cole  PD, Beckwith  KA, Vijayanathan  V, Roychowdhury  S, Smith  AK, Kamen  BA.  Folate homeostasis in cerebrospinal fluid during therapy for acute lymphoblastic leukemia.  Pediatr Neurol. 2009;40(1):34-41.PubMedGoogle ScholarCrossref
9.
Caron  JE, Krull  KR, Hockenberry  M, Jain  N, Kaemingk  K, Moore  IM.  Oxidative stress and executive function in children receiving chemotherapy for acute lymphoblastic leukemia.  Pediatr Blood Cancer. 2009;53(4):551-556.PubMedGoogle ScholarCrossref
10.
Krull  KR, Bhojwani  D, Conklin  HM,  et al.  Genetic mediators of neurocognitive outcomes in survivors of childhood acute lymphoblastic leukemia.  J Clin Oncol. 2013;31(17):2182-2188.PubMedGoogle ScholarCrossref
11.
Pui  CH, Campana  D, Pei  D,  et al.  Treating childhood acute lymphoblastic leukemia without cranial irradiation.  N Engl J Med. 2009;360(26):2730-2741.PubMedGoogle ScholarCrossref
12.
Pui  C-H.  Toward optimal central nervous system-directed treatment in childhood acute lymphoblastic leukemia.  J Clin Oncol. 2003;21(2):179-181.PubMedGoogle ScholarCrossref
13.
Bhojwani  D, Sabin  ND, Pei  D,  et al.  Methotrexate-induced neurotoxicity and leukoencephalopathy in childhood acute lymphoblastic leukemia.  J Clin Oncol. 2014;32(9):949-959.PubMedGoogle ScholarCrossref
14.
Kishi  S, Griener  J, Cheng  C,  et al.  Homocysteine, pharmacogenetics, and neurotoxicity in children with leukemia.  J Clin Oncol. 2003;21(16):3084-3091.PubMedGoogle ScholarCrossref
15.
National Institutes of Health; National Cancer Institute. Common Terminology Criteria for Adverse Events (CTCAE), Version 4.03. https://evs.nci.nih.gov/ftp1/CTCAE/CTCAE_4.03_2010-06-14_QuickReference_8.5x11.pdf. June 14, 2010. Accessed August 1, 2016.
16.
Benjamini  Y, Hochberg  Y.  Controlling the false discovery rate: a practical and powerful approach to multiple testing.  J R Stat Soc B. 1995;57(1):289-300.Google Scholar
17.
Rodriguez  JI, Kern  JK.  Evidence of microglial activation in autism and its possible role in brain underconnectivity.  Neuron Glia Biol. 2011;7(2-4):205-213.PubMedGoogle ScholarCrossref
18.
Bond  J, Hough  R, Moppett  J, Vora  A, Mitchell  C, Goulden  N.  ‘Stroke-like syndrome’ caused by intrathecal methotrexate in patients treated during the UKALL 2003 trial.  Leukemia. 2013;27(4):954-956.PubMedGoogle ScholarCrossref
19.
Villoslada  P, Genain  CP.  Role of nerve growth factor and other trophic factors in brain inflammation.  Prog Brain Res. 2004;146:403-414.PubMedGoogle ScholarCrossref
20.
Jiang  R, Brummett  BH, Babyak  MA, Siegler  IC, Williams  RB.  Brain-derived neurotrophic factor (BDNF) Val66Met and adulthood chronic stress interact to affect depressive symptoms.  J Psychiatr Res. 2013;47(2):233-239.PubMedGoogle ScholarCrossref
21.
Grabe  HJ, Schwahn  C, Mahler  J,  et al.  Genetic epistasis between the brain-derived neurotrophic factor Val66Met polymorphism and the 5-HTT promoter polymorphism moderates the susceptibility to depressive disorders after childhood abuse.  Prog Neuropsychopharmacol Biol Psychiatry. 2012;36(2):264-270.PubMedGoogle ScholarCrossref
22.
Yang  M, Moon  C.  Neurotoxicity of cancer chemotherapy.  Neural Regen Res. 2013;8(17):1606-1614.PubMedGoogle Scholar
23.
Wilson  CJ, Finch  CE, Cohen  HJ.  Cytokines and cognition–the case for a head-to-toe inflammatory paradigm.  J Am Geriatr Soc. 2002;50(12):2041-2056.PubMedGoogle ScholarCrossref
24.
Abbott  NJ, Rönnbäck  L, Hansson  E.  Astrocyte-endothelial interactions at the blood-brain barrier.  Nat Rev Neurosci. 2006;7(1):41-53.PubMedGoogle ScholarCrossref
25.
Wardill  HR, Mander  KA, Van Sebille  YZA,  et al.  Cytokine-mediated blood brain barrier disruption as a conduit for cancer/chemotherapy-associated neurotoxicity and cognitive dysfunction.  Int J Cancer. 2016;139(12):2635-2645.PubMedGoogle ScholarCrossref
26.
Savitz  J, Solms  M, Ramesar  R.  The molecular genetics of cognition: dopamine, COMT and BDNF.  Genes Brain Behav. 2006;5(4):311-328.PubMedGoogle ScholarCrossref
×