Identifying patterns of neurocognitive dysfunction through direct comparison of children with leukemia, central nervous system tumors, and sickle cell disease

Abstract

Purpose:

To quantify and compare the magnitude and type of neurocognitive dysfunction in at-risk children with central nervous system (CNS) tumors, acute lymphoblastic leukemia (ALL), and sickle cell disease (SCD) using a common instrument and metric to directly compare these groups with each other.

Methods:

Fifty-three participants between the ages of 7 and 12 years (n = 27 ALL, n = 11 CNS tumor, n = 15 SCD) were enrolled and assessed using the NIH Toolbox Cognition Battery (NIHTCB). Participants with ALL or CNS tumor were 0–18 months posttherapy, while participants with SCD possessed the SS or Sβ⁰ genotype, took hydroxyurea, and had no known history of stroke.

Results:

Independent sample t-tests showed that participants with ALL and CNS tumor experienced greatest deficits in processing speed (ALL d=−0.96; CNS tumor d=−1.2) and inhibitory control and attention (ALL d =−0.53; CNS tumor d =−0.97) when compared with NIHTCB normative data. Participants with SCD experienced deficits in cognitive flexibility only (d =−0.53). Episodic memory was relatively spared in all groups (d =−0.03 to −0.32). There were no significant differences in function when groups were compared directly with each other by analysis of variance.

Conclusions:

Use of a common metric to quantify the magnitude and type of neurocognitive dysfunction across at-risk groups of participants by disease shows that participants perform below age-expected norms in multiple domains and experience dysfunction differently than one another. This approach highlights patterns of dysfunction that can inform disease- and domain-specific interventions.

1 |. INTRODUCTION

It is increasingly evident that brain development and cognitive function are impacted in a variety of childhood illnesses as a result of disease, treatment, and social-environmental factors.¹ Neurocognitive difficulties are a significant problem with insidious and far-reaching effects in the daily lives of children; difficulties in executive function in particular can impact school performance, occupational attainment, and quality of life for both children and their families.² For the pediatric hematologist oncologist, the groups most impacted are childhood cancer survivors and children with sickle cell disease (SCD).^2–6

Among childhood cancer survivors, those with acute lymphoblastic leukemia (ALL) and central nervous system (CNS) tumors appear to be most severely impacted.^2–5,7 Meta-analyses have shown that pediatric ALL survivors, despite eliminating cranial radiation therapy (CRT) for most patients in the majority of modern protocols, continue to perform worse than their healthy peers with full-scale intelligence quotients (FSIQ) in the 25–30th percentile and specific deficits in the domains of working memory, executive function, and attention when assessed in long term follow up.^4,8 Demographic risk factors such as female sex, age under 10 years, and treatment with high-dose methotrexate are thought to exacerbate this risk for dysfunction.^10,2,9 Survivors of CNS tumors fare even worse in the long term, with FSIQ below the 22nd percentile and specific deficits in domains of processing speed, working memory, executive function, and attention.^11,3,5 Risk factors are demographic, tumor, and treatment related, including young age at diagnosis and treatment, presence of hydrocephalus, tumor location, presence of cerebellar mutism syndrome, and dose and extent of radiation field.³ School-aged children with SCD similarly perform at levels below healthy peers, with FSIQ below the 22nd percentile and executive function at the 19th percentile, with increased risk conferred by disease complications such as the presence of silent cerebral infarct, stroke, and increased transcranial Doppler velocities.^12,6

Current knowledge about neurocognitive function in these groups of children comes from studying these groups separately, with a number of different measures across studies, and comparing these patients’ performance against normative data, healthy siblings, or healthy controls. While this type of comparison is useful for calculating the magnitude of dysfunction relative to healthy peers, it does not directly provide insight into how children with ALL, CNS tumors, and SCD are performing relative to each other. Employment of a common measure of neuropsychological function allows for direct comparison between at-risk groups of patients. The present study enrolled the three disease groups at greatest risk for dysfunction and measured neurocognitive function using a common metric, allowing us to put these groups on the same “measuring stick.” Placement on the same “measuring stick” is useful for several reasons: first, it lends an additional perspective in identifying the highest risk patients and triaging which patients should receive neuropsychological screening and assessment early; second, it highlights whether cognitive function is differentially impacted in different disease groups. If groups experience similar patterns of dysfunction across domains, common interventions might be developed and applied across disease groups. If, however, unique patterns of cognitive function emerge, interventions should be tailored to be both domain and disease specific. Our primary aim was to quantify the magnitude and pattern of neurocognitive dysfunction in at-risk disease groups (i.e., ALL, CNS tumors, and SCD) by measuring function with a common metric, the National Institutes of Health Toolbox Cognition Battery (NIHTCB),¹³ to allow for direct comparison between groups. The NIHTCB was chosen for ease of administration, inclusive age range, and multiple domains, particularly within executive function, included in its battery.

First, we considered the three groups together by collapsing across diagnoses and directly comparing cognitive function in the present sample to normative data. Next, to put our findings in the context of prior literature, we separated participants into groups by diagnosis and compared cognitive function to normative data. We hypothesized that participants with ALL, CNS tumors, and SCD would demonstrate significantly worse neurocognitive function in all domains assessed when compared with normative data, with participants with ALL demonstrating deficits with small to medium effects, and participants with CNS tumors or SCD demonstrating deficits with medium to large effect sizes. Next, when comparing these groups directly to one another, we hypothesized that participants with ALL would show the smallest deficits in domains of neurocognitive function. Participants with CNS tumors would show the largest deficits in discrete domains of neurocognitive function, with participants with SCD performing in between the other two disease groups.

2 |. METHODS

The present study employed a cross-sectional design to measure and directly compare neurocognitive function among participants with ALL, CNS tumors, and SCD. The NIHTCB was selected as a common metric to facilitate direct comparison between groups of participants in order to quantify the magnitude and pattern of neurocognitive deficits in these at-risk populations.

2.1 |. Participants

Eligible participants were identified through institutional database review, physician or nurse referral, appointment review, and departmental meetings. Study activities were approved by the Colorado Multiple Institutional Review Board, and informed consent and assent were obtained prior to study participation.

Strict inclusion/exclusion criteria were applied in an effort to make groups as homogeneous as possible, knowing a significant amount of heterogeneity would be present between groups. Eligible participants were between the ages of 7 and 12 years with ALL, CNS tumors, or SCD, were English or Spanish speaking, and did not have a comorbid neurologic disorder (e.g., neurofibromatosis) or known intellectual disability. Participants with ALL or CNS tumors must have completed therapy within 18 months to be eligible. Most children with SCD are diagnosed at birth, thus this criterion did not apply. Children with CNS tumors were eligible if diagnosed by biopsy or tumor markers. Participants with SCD were eligible for participation if they took hydroxyurea and possessed the SS or Sβ⁰ genotype. Participants with SCD were ineligible if they had a history of overt stroke, received chronic exchange transfusions, or had undergone bone marrow transplantation. Among participants with ALL, those with B cell lineage were eligible, while those with a history of symptomatic hyperleukocytosis were ineligible.

2.2 |. Measures

Neurocognitive function was measured with the NIHTCB, an iPad-based assessment of cognitive abilities for ages 3–85 years.¹³ Cognitive measures included three tasks of executive function (including List Sorting Working Memory Test to assess working memory, Flanker Inhibitory Control and Attention Test to measure inhibitory control and attention, and Dimensional Change Card Sort Test to assess cognitive flexibility) as well as measures of processing speed (Pattern Comparison Processing Speed Test) and episodic memory (Picture Sequence Memory Test). All five domains were taken together to yield a cognitive fluid composite score. Neurocognitive results are reported as age-adjusted standard scores with a mean of 100 and standard deviation of 15. The measure was chosen for ease of administration, inclusive age range, and for the multiple domains encompassed in its battery. NIHTCB was administered in either the English or Spanish version, depending on the preference of the participant. Spanish-speaking participants were additionally accompanied by an in-person interpreter to facilitate administration. In the setting of the COVID-19 pandemic, research visits were scheduled concurrently with previously scheduled clinical appointments. Children and their caregivers were offered compensation for study enrollment. Upon completion of neurocognitive assessment, families were provided letters summarizing their child’s performance. Results disclosure was coordinated with participants’ primary hematologist or oncologist, as well as psychologists to assist in determining next steps in evaluation, if needed.

Clinically relevant disease and treatment data were extracted from medical records of enrolled participants. Examples include diagnoses, comorbidities, treatment received, CNS tumor location, baseline hemoglobin for participants with SCD, and CNS involvement in participants with ALL.

2.3 |. Statistical analyses

Demographic and clinical variables were compared between the three disease groups with Chi Squared analysis or independent samples t-tests, depending on variable type.

First, to quantify the magnitude and pattern of cognitive dysfunction in participants with ALL, CNS tumors, and SCD relative to normative data, a series of independent two-tailed one sample t-tests were performed. Effect sizes are reported using Cohen’s d statistic; 0.2 is considered small, 0.5 is considered medium, and 0.8 is a large effect.¹⁴ Second, groups of participants were compared directly with each other using a one-way analysis of variance (ANOVA) to determine if these patient groups differ from one another. If the overall F-test was significant, pairwise differences were planned using t-tests. Multiple comparison adjustment was done using the Bonferroni correction. Last, given the difficulties associated with face-to-face NIHTCB administration in the midst of the COVID-19 pandemic, feasibility and acceptability are discussed in the context of percentage of participants enrolled of total eligible patients.

3 |. RESULTS

Between May 2021 and June 2022, 92 participants were identified as eligible for enrollment (ALL = 41, CNS tumor = 29, SCD = 22). One child moved out of state, seven participants declined or withdrew their participation, three aged out, and 19 were unreachable or unavailable. Among participants with ALL or CNS tumors, eight surpassed the 18-month mark since completing therapy and three relapsed. Ultimately 53 participants were enrolled and participated in neurocognitive assessment, including 27 participants with ALL (66% eligible), 11 participants with CNS tumors (38%), and 15 with SCD (68%). Complete cognitive testing data is available on 50 participants; three participants are missing scores on one of the five sub-tests due to technical errors in the NIHTCB scoring program calculating age-adjusted scores or participant preference.

Demographic information is presented in Table 1, and clinical information is presented in Table 2. Participants with ALL were expected to be significantly younger at diagnosis than their CNS counterparts, given the duration of ALL therapy. Similarly, differences in race and ethnicity between groups were expected given the increased frequency of sickle cell gene variants in the Black population and the increased incidence of ALL in Hispanic children.^15,16

TABLE 1.

Demographic characteristics by diagnosis.

	ALL		CNS tumors		SCD
	N (%)	M (SD)	N (%)	M (SD)	N (%)	M (SD)	p-value
Sex							.023
Female	10 (37%)	-	9 (82%)	-	5 (33%)	-
Male	17 (63%)	-	2 (18%)	-	10 (67%)	-
Current age	-	9.2 (1.7)	-	9.7 (1.5)	-	8.1 (1.6)	.043
Age at diagnosis	-	5.8 (2.0)	-	8.6 (1.6)	-	-	<.001
Months off-therapy	-	5.4 (5.0)	-	7.2 (1.8)	-	-	.365
Race							<.001
Black	1 (3.7%)	-	0	-	13 (87%)	-
White	19 (70%)	-	9 (82%)	-	0	-
Hawaiian/Pacific Islander	1 (3.7%)	-	0	-	0	-
Native American/Alaska Native	1 (3.7%)	-	0	-	0	-
Mixed Race	0	-	0	-	2 (13%)	-
Ethnicity							<.001
Hispanic or Latinx	5 (19%)	-	3 (27%)	-	0	-
Non-Hispanic or Latinx	17 (63%)	-	7 (64%)	-	15 (100%)	-
Language							.496
English-only	23 (85%)	-	10 (91%)	-	15 (100%)	-
Spanish-only	2 (7%)	-	0	-	0	-
Bilingual	2 (7%)	-	1 (9%)	-	0	-

ALL, acute lymphoblastic leukemia; CNS, central nervous system; SCD, sickle cell disease.

TABLE 2.

Clinical characteristics by diagnosis.

ALL		CNS tumors		SCD
	N (%)		N (%)		N (%)	M(SD)
CNS status		Tumor type		Genotype
CNS 1	27(100%)	High grade glioma	1 (9%)	SS	14 (93%)	-
CNS 2/3	0	Low grade glioma	1 (9%)	Sβ0	1 (7%)	-
HD-MTX	0	Medulloblastoma/embryonal	2 (18%)	HgbA	-	8.7 g/dL (1.3)
MTX toxicity	7 (26%)	Ependymoma	1 (9%)	HgbF	-	22.5 (11.9)
Prolonged ICU stay	0	Germ cell tumor	3 (27%)	SCI	0	-
Mean doses IT chemotherapy	1 (4%)	Craniopharyngioma	1 (9%)	Recent TCD
		Other	2 (18%)	Normal	12 (80%)	-
		Tumor location		Conditional	3 (20%)	-
		Supratentorial	7 (63%)	Abnormal	0	-
		Infratentorial	4 (36%)	Number hospitalizations	-	4.5 (4.7)
		Treatment received
		Surgery	9 (82%)
		Chemotherapy	7 (64%)
		Radiation therapy	10(91%)
		Hydrocephalus	6(11%)

Prolonged ICU stay defined as ≥72 h. ALL, acute lymphoblastic leukemia; CNS, central nervous system; HD-MTX, high dose methotrexate; Hgb, hemoglobin; ICU, intensive care unit; IT, intrathecal; SCD, sickle cell disease; SCI, silent cerebral infarct; TCD, transcranial Doppler.

When collapsed across disease groups and compared with normative data, the sample as a whole performed below the normative mean of 100, with significant differences detected in overall fluid composite score (d =−0.75), working memory (d =−0.3), processing speed (d =−0.8), inhibitory control and attention (d =−0.55), and cognitive flexibility (d =−0.49), with episodic memory remaining relatively spared (d =−0.17; Table 3).

TABLE 3.

NIH Toolbox cognition battery age-adjusted standard scores compared with normative data.

	Entire sample			ALL			CNS tumor			SCD
	M	SD	p-value*	M	SD	p-value*	M	SD	p-value*	M	SD	p-value*
Working memory	95.5 N = 52	(13.3)	.033	93.4 N = 26	(14.5)	.026	98.5 N = 11	(13.2)	.741	97.1 N = 15	(11.1)	.440
Processing speed	88.0 N = 51	(20.8)	<.001	85.6 N = 25	(19.0)	<.001	82 N = 11	(21.7)	<.001	96.3 N = 15	(22.0)	.759
Episodic memory	97.4 N = 51	(14.2)	.327	99.5 N = 26	(15.3)	.866	95.2 N = 11	(12.6)	.290	95.1 N = 14	(13.3)	.488
Inhibitory control and attention	91.8 N = 52	(13.9)	<.001	92.1 N = 26	(14.4)	.008	85.4 N = 11	(10.3)	.001	96.0 N = 15	(14.5)	.342
Cognitive flexibility	92.7 N = 52	(14.0)	<.001	93.4 N = 26	(11.7)	.026	91.7 N = 11	(15.9)	.068	92.3 N = 15	(17.1)	.043
Cognition fluid composite	88.7 N = 50	(14.1)	<.001	88.8 N = 25	(12.9)	<.001	84.1 N = 11	(16.3)	<.001	92.1 N = 14	(14.2)	.118

ALL, acute lymphoblastic leukemia; CNS, central nervous system; SCD = sickle cell disease.

^*Significance level after Bonferroni correction for multiple comparisons p ≤ .0016.

When separated by diagnosis and compared with normative data, cognitive fluid composite scores were significantly below normative means in participants with ALL (d=−0.75) and CNS tumors (d=−1.06). Though the composite score was below the mean in participants with SCD, it did not reach statistical significance (d =−0.53). When participants with ALL were compared with normative data, significant differences were noted in four of the five domains assessed, with medium negative effects noted in inhibitory control and attention (d =−0.53) and large negative effects noted in processing speed (d =−0.96; Table 3 and Figure 1). When participants with CNS tumors were compared with normative data, similar domains were affected, with significant deficits in processing speed (d =−1.2) and inhibitory control and attention (d =−0.97) being most profound. Participants with SCD generally did not demonstrate differences that reached statistical significance when compared with normative data but did demonstrate a negative medium effect in the domain of cognitive flexibility (d =−0.51). When the three disease groups were compared directly with each other using ANOVA, no significant differences were elucidated.

FIGURE 1.

Cognitive performance by domain and disease group. p-values represent independent samples t-tests comparing to normative data. Error bars represent 1 SD.

4 |. DISCUSSION

This study reports on the magnitude and type of neurocognitive dysfunction in at-risk disease groups in the pediatric hematology oncology population (i.e., ALL, CNS tumors, and SCD) by measuring function with a common metric to allow for direct comparison between groups. Overall, our findings demonstrate that when a common metric is employed (in this case, the NIHTCB), children perform below age-expected norms in multiple domains and experience different types of neurocognitive dysfunction based on disease group.

The present study is the first of its kind to simultaneously enroll the three pediatric hematology-oncology disease groups at highest risk for neurocognitive dysfunction. When collapsed across diagnoses and considered as a whole, participants experienced deficits in multiple domains, including working memory, processing speed, inhibitory control and attention, and cognitive flexibility. Interestingly, these groups generally demonstrated preserved episodic memory.

Participants with ALL experienced greater deficits than anticipated. While we predicted participants would exhibit difficulties with small to medium effect sizes, the current sample demonstrated effect sizes approaching medium to large. The most profound deficits were noted in processing speed and inhibitory control and attention, with a smaller fraction of the sample experiencing difficulty in working memory and cognitive flexibility. The large effect size in processing speed in our cohort is larger than what is reported in the literature, though the current sample was assessed at an average of 5.4 months posttherapy, much more proximal to completion of therapy than is generally performed in cross-sectional studies.¹⁷ This difference in effect size magnitude at different times of assessment raises the question of whether children experience gains in processing speed the farther out they are from treatment. A recent longitudinal study examined this trajectory and found that males and participants treated with standard- or high-risk therapies were at increased risk for poorer processing speed at 2 years posttreatment, but the sample as a whole performed in the average range at the end of therapy with no significant change over time to 2 years post.¹⁸ This is somewhat at odds with the body of work that has demonstrated deficits in processing speed in long term survivors of ALL.^19–22,9

In keeping with prior studies examining attention among survivors of ALL,^18,20,21 a medium effect in the negative direction was noted in the domain of inhibitory control and attention in the present sample. Deficits in attention near end of therapy are noteworthy, given their contribution to the development of executive function over time, and the association of executive function with quality of life.^18,19,23

Among participants with CNS tumors, deficits were largely aligned with what was hypothesized. Processing speed and inhibitory control and attention were most profoundly impacted, and a smaller proportion of participants exhibited medium effects in cognitive flexibility. It is noteworthy that participants with CNS tumor or ALL demonstrated dysfunction in similar cognitive domains, though participants with CNS tumors experienced deficits with greater magnitude than the ALL cohort; these more profound deficits did not reach statistical significance, however, likely due to small sample size in the CNS tumor cohort. Deficits in the domains of processing speed and inhibitory control and attention had large, negative effect sizes, in keeping with what has previously been described.^11,24,5 What is noteworthy about the current sample is the relative preservation of working memory and episodic memory, contrary to previous reports.^11,24

Participants with SCD generally performed in the average range, with the exception of cognitive flexibility. This was contrary to our hypothesis that participants would exhibit deficits with medium to large effect sizes in multiple domains of cognition. The medium, negative effect seen in cognitive flexibility is similar to patterns previously seen for executive function,⁶ but performance in the average range in other domains was unanticipated. While the current sample had no known SCI, prior meta-analyses suggest that children without history of SCI generally perform below their healthy peers.^25,26,6 It is possible that participants at our center with poorer cognitive function were made ineligible for the study because of strict inclusion/exclusion criteria (e.g., hydroxyurea requirement, no history of overt stroke, no current chronic exchange transfusions). There is evidence that early administration of hydroxyurea correlates with preserved function in several domains of cognition.^27,28 Prior meta-analyses have included heterogeneous samples of patients taking hydroxyurea as well as those not taking it, which may account for poorer cognitive function quantified in these analyses.^12,6 Not only was the current sample prescribed hydroxyurea, but the mean fetal hemoglobin (22.5%) suggests the present sample was adherent to this regimen.

When compared directly with each other, there were no significant differences found between the three groups. While we hypothesized that participants with ALL would demonstrate the fewest deficits, followed by participants with SCD, with participants with CNS tumors evidencing the most profound dysfunction, our findings demonstrated that, when compared with normative data, participants with SCD had relatively preserved function (save for cognitive flexibility), followed by participants with ALL, with participants with CNS tumors exhibiting the most profound deficits.

It is notable that there were no statistically significant differences in episodic memory across each of the disease groups. Previous work estimates episodic memory performance in the average range among participants with ALL treated with chemotherapy-only protocols and participants with SCD.^20,29 However, participants with CNS tumors have historically experienced deficits with large effect sizes that are made worse by CRT and seizure disorders.^11,24 While it is notable that our cohort of participants with CNS tumors exhibited episodic memory deficits with small magnitude (d =−0.32), the small sample size of the current CNS tumor cohort likely limits our power to detect a statistically significant difference compared with normative data or to other disease groups. One hypothesis that may explain why episodic memory is relatively preserved has to do with the developmental trajectory of brain regions underlying this skill. It is plausible that the hippocampus and medial temporal lobe, home to episodic memory, are less affected by adjuvant therapies because they complete myelination earlier than structures in the frontal cortex that house executive functions (e.g., working memory, attentional control and inhibition, cognitive flexibility) that undergo myelination and pruning into adulthood.^30,31 This developmental theory does not entirely explain the current findings, however, given the body of literature describing hippocampal injury by chemotherapy and radiation in both pediatric and adult populations.^32,33

Use of the NIHTCB to facilitate direct comparison between these at-risk groups of participants was determined to be acceptable and feasible among participants with ALL and SCD, with 66–68% of eligible participants enrolling onto the study, even through the COVID-19 pandemic. This battery was developed to provide a brief, reliable assessment of subdomains of cognition that impact development, education, and employment; establishing its acceptability and feasibility in clinical samples is of utmost importance. Use in the present CNS tumor sample was less feasible, however, with only 38% of eligible participants enrolling onto the study. Testing in this group was made difficult by less frequent clinic visits, greater number of appointments per day in the outpatient setting, and increased frequency of sedation for surveillance imaging. Nevertheless, successful use overall of this battery among participants with cancer and blood disorders demonstrates its efficiency, acceptability, and feasibility.

This study was the first of its kind to enroll children at greatest risk for neurocognitive dysfunction across disease groups in the pediatric hematology oncology population. Employing a common methodology and common metric across the disease groups allows for identification of disease- and domain-specific patterns of dysfunction, paving the way for development of disease- and domain-specific interventions.

This study is not without limitations. First, the sample size, particularly of the CNS tumor cohort, was relatively small and contributed to being underpowered for direct comparison between the three groups. Additionally, small sample size in the CNS tumor cohort limits generalizability to the larger population of CNS tumor survivors. Second, because neurocognitive assessment was done in the context of a busy clinic day to limit time in the hospital in the midst of the COVID-19 pandemic, scores may have been impacted by fatigue and/or anxiety. Third, the narrow inclusion/exclusion criteria, particularly in the SCD cohort, may have inflated neurocognitive performance and may limit generalizability of the present findings to other groups with ALL, CNS tumor, and SCD. Fourth, the onset and evolution of neurocognitive dysfunction varies across the three disease groups, and the cross-sectional nature of the current study limits our ability to capture comparisons at different time points. Finally, differences in race and ethnicity between the three groups were not accounted for in analyses since they were highly correlated with diagnosis. Future studies should aim to utilize these emerging patterns of neurocognitive dysfunction to develop disease- and domain-specific interventions. In addition, focus on potential shared risk factors affecting all disease groups, namely social determinants of health, should be prioritized to build more robust risk prediction models for neurocognitive dysfunction among children with cancer and blood disorders.

Abbreviations:

ALL

acute lymphoblastic leukemia

ANOVA

analysis of variance

CNS

central nervous system

CRT

cranial radiation therapy

FSIQ

full-scale intelligence quotient

NIHTCB

National Institutes of Health Toolbox Cognition Battery

SCD

sickle cell disease