Abstract

The prevalence of late effects following allogeneic hematopoietic cell transplantation (HCT), a curative treatment for pediatric leukemia, is high: 79% of HCT recipients experience chronic medical conditions. The few extant studies of cognitive late effects have focused on intelligence and are equivocal about HCT neurotoxicity. In an archival study of 30 children (mean transplant age = 6 years), we characterize neuropsychological predictors of academic outcomes. Mean intellectual and academic abilities were average, but evidenced extreme variability, particularly on measures of attention and memory: ∼25% of the sample exhibited borderline performance or lower. Medical predictors of outcome revealed paradoxically better memory associated with more severe acute graft-versus-host disease (GVHD) and associated with steroid treatment. Processing speed and memory accounted for 69% and 61% of variance in mathematics and reading outcomes, respectively. Thus, our findings revealed neurocognitive areas of vulnerability in processing speed and memory following HCT that contribute to subsequent academic difficulties.

Introduction

Allogeneic hematopoietic cell transplantation (HCT) in pediatric patients is a curative option for a variety of pediatric disorders including malignancies, disorders of the hematopoietic system, immunologic disorders, and metabolic syndromes. Treatment-related mortality has declined secondary to advances in human leukocyte antigen matching specificity, graft-versus-host disease (GVHD) prevention and treatment, reductions in intensity of conditioning regimens, and improved supportive care measures for transplant recipients (Bejanyan et al., 2015; Pingali & Champlin, 2015; Rasche, Kapp, Einsele, & Mielke, 2014; Worel et al., 2002). However, as our ability to cure pediatric patients with HCT improves, a growing population of transplant survivors has emerged with chronic medical conditions related to their curative therapy. The scope of chronic health conditions is broad and common systems affected include endocrine, cardiac, pulmonary, genitourinary, immunologic, musculoskeletal, neuropsychiatric, and the development of secondary malignancies (Abou-Mourad et al., 2010; Armenian et al., 2011; Ferry et al., 2007; Sun et al., 2010). The prevalence of late effects from therapy is high, with a recent analysis from the Bone Marrow Transplant Survival Study documenting a chronic medical condition in 79% of pediatric HCT recipients and 26% of patients suffering from a severe or life threatening medical ailment (Armenian et al., 2011). The prevalence of late effects is greater than that in pediatric patients who undergo chemotherapy alone. Unique aspects of HCT that place patients at increased risk of chronic medical problems include intensity of chemotherapy, the inclusion of total body irradiation (TBI) in therapy (Bejanyan et al., 2015; Pingali & Champlin, 2015), inflammation related to stem cell engraftment and GVHD, and toxicities related to short- and long-term immune suppression (Rasche et al., 2014; Worel et al., 2002).

Given the increasing population of transplant survivors and the high prevalence of morbidity despite cure of the original disease, a great deal of effort is now being devoted to characterizing late complications and identifying strategies to detect and possibly prevent late medical problems. Cardiac, endocrine, ocular, and musculoskeletal complications have been well defined and interventions are now being employed to decrease the occurrence of disease in these systems. One area of transplant late effects medicine, however, that has not been well characterized is the effect on cognition in pediatric transplant patients.

Only a small number of studies have reported on cognitive functioning in pediatric transplant recipients, and the majority of these studies have focused on global measures of intelligence. In addition, the studies have reported inconsistent findings among tested patients about potential neurotoxicity of HCT. No significant changes in mean Full Scale, Verbal, or Performance Intellectual Quotient (IQ) were found in a prospective study by Simms, Kazak, Golomb, Goldwein, and Bunin (2002) or in a prospective study 1 or 2 years post-transplant by Kupst and colleagues (2002). Surprisingly, Barrera and Atenafu (2008) reported that transplant survivors actually had a statistically higher Performance IQ at 1 and 2 years post-transplant and a higher Full-Scale IQ at 2 years post-transplant when compared with a normative population. In contrast, Kramer, Crittenden, DeSantes, and Cowan (1997) identified a significant drop in mean IQ between baseline and 1 year as well as declines in adaptive functioning in young children undergoing transplant.

When examining academics, both the Kupst and Barrera studies did detect deficits, particularly in areas of mathematics. Armstrong (2001) also identified deficits in mathematics and reading post-transplant as common areas of concern. These academic deficits may be related to absence from school and consequent reduced ability to gain early operational skills as opposed to direct neurotoxicity of treatment. An equally plausible consideration is that subtle inefficiencies in core cognitive processes, such as processing speed and memory, disrupt the ongoing acquisition of academic skills.

Indeed, when researchers focused on individual aspects of cognition, poorer performance on several neuropsychological domains emerged. Shah and colleagues (2008) reported declines in Verbal and Performance IQ, visual motor skills, and long-term memory in patients who were 5 years post-transplant. Armstrong's (2001) review also identified problems in processing speed, attention, and memory for visual information post-transplant. Despite reporting average IQ in transplant survivors, Perkins and colleagues (2007) documented significant deficits in multiple domains including visual motor skills, fine motor speed, sustained attention, and adaptive behavior.

Missing from the literature is an analysis of relationships between core microlevel cognitive processes and higher order macrolevel academic outcomes. Therefore, the current study sought to characterize the microlevel patterns of processing speed, attention, and memory in relation to macrolevel academic functioning post-HCT. As such, our goal was to determine whether processing speed and memory could account for the variance in outcome in academic performance. Processing speed and memory were considered a priori as potential predictors given the aforementioned deficits noted in pediatric HCT survivors and given that decreased processing speed and memory are known to contribute to poor academic achievement in a wide range of pediatric medical conditions. Long-term deficits in processing speed are reported in children with a history of low birth weight (Murray, Scratch, Thompson, Inder, & Doyle, 2014), traumatic brain injury (Anderson, Godfrey, Rosenfeld, & Catroppa, 2012), and following exposure to teratogens such as alcohol (e.g., fetal alcohol spectrum disorders; Kodituwakku, 2007). Similarly, long-term inefficiencies and impairments in memory have been documented in these conditions (Arnett et al., 2013; Murray et al., 2014) and described in comprehensive reviews (Kodituwakku, 2007). Indeed, deficits in processing speed (Schatz, Kramer, Ablin, & Matthay, 2000) and memory (Peterson et al., 2008) have been reported following both cranial radiation therapy and chemotherapy, respectively, for acute lymphoblastic leukemia (ALL). The aforementioned studies suggest that a complex relationship exists between inefficiencies in speed of information processing and academic outcomes. We hypothesized (1) our sample would demonstrate deficits in processing speed and memory consistent with Shah and colleagues (2008) and Perkins and colleagues (2007) as well as (2) poorer overall academic functioning in reading and mathematics. We further hypothesized (3) variance in reading and mathematics performance would be largely accounted for by these process variable inefficiencies. Given limited investigations from which to extrapolate hypotheses regarding biomedical factors influencing outcomes, we sought to investigate these potential predictors through exploratory analyses. Specifically, both mediator and moderator models were tested with GVHD and steroid treatment as predictors, mediators, and moderators of academic outcome.

Methods

Participants

Participants for this study were obtained through an archival review of the Pediatric Hematopoietic Cell Transplant Clinic and the Neuropsychology Section at the University of Michigan Health Systems. The records of 30 children who had received an allogeneic HCT at <18 years of age and for whom at least one comprehensive neuropsychological assessment had been conducted post-transplant were included. Participants were excluded from the study if they had a known history of a neurodevelopmental or neurological disorder (e.g., epilepsy) that predated the onset of cancer, or traumatic brain injury with a loss of consciousness or severe central nervous system (CNS) infection either pre- or postillness. The protocol and all study procedures were approved by the University of Michigan Health Systems Institutional Review Board.

During the period between 1998 and 2011, 239 pediatric patients were cared for through the Pediatric Hematopoietic Cell Transplant Clinic. To participate in the study, patients had to be 1-year post-HCT and at least 4 years old, leaving only 66% (n = 157) of the population eligible for the study. Of these, 19% (n = 30) completed a comprehensive assessment of neuropsychological and psychosocial functioning in the Neuropsychology Section in the Division of Psychiatry. This represents a total of 12.5% of children who underwent HCT during the 13-year period and does not correct for mortality. Reason for referral was dictated by the clinical practice of the patient's transplant physician and the ability to complete the testing based on the health of the participant.

Materials and Procedure

Medical Data

Medical data obtained through an existing database included (i) participant demographic information (gender, age of HCT, and time since HCT), (ii) underlying diagnosis and status at transplant, (iii) donor relation and degree of matching, (iv) stem cell source, (v) conditioning regimen, (vi) GVHD prophylaxis, (vii) existence and grading of acute GVHD (aGVHD), and (viii) steroid treatment for aGVHD. Additional medical data obtained through a retrospective chart review included information regarding (ix) existence and severity of chronic GVHD (cGVHD) and (x) steroid treatment for cGVHD. These served as our independent variables.

Neuropsychological Functioning

The neuropsychological data included in this investigation included (i) demographic variables (i.e., gender, age at testing, and handedness), (ii) intellectual functioning, (iii) academic achievement, (iv) attention, (v) memory, and (vi) processing speed. See Table 1 for clarification of the domains and instruments used in the study.

Table 1.

Descriptives and outcomes of neuropsychological tests

Domain Test Standard Score M (SDRange 
Intelligence Wechsler Intelligence Scale for Children IV (Wechsler, 2003) (WISC-IV): Compositesa   
Verbal Comprehension Index 94 (12.78) 67–121 
Perceptual Reasoning Index 90 (12.22) 65–112 
Working Memory Index 90 (15.02) 52–120 
Processing Speed Index 88 (15.85) 50–116 
Full-Scale IQ 87 (16.34) 50–115 
Wechsler Intelligence Scale for Children IV (WISC-IV): Subtestsb   
Similarities 10 (2.18) 5–13 
Vocabulary 9 (2.52) 3–13 
Comprehension 9 (2.63) 3–14 
Block design 7 (2.59) 3–14 
Picture concepts 9 (2.63) 2–13 
Matrix reasoning 9 (2.95) 3–16 
Digit span 8 (2.63) 2–13 
Letter number sequencing 9 (3.15) 1–14 
Coding 8 (3.13) 1–14 
Symbol search 8 (3.22) 1–13 
Academic achievement Woodcock–Johnson III Tests of Achievement (Woodcock, McGrew, & Mather, 2007) (WJ III ACH)a   
Math composite 91 (21.71) 45–124 
Reading composite 91 (22.34) 27–143 
Writing composite 97 (14.81) 72–130 
Letterword 100 (15.98) 79–148 
Passage Comprehension 92 (15.60) 65–138 
Calculation 93 (19.19) 57–124 
Spelling 97 (17.33) 68–130 
Writing sample 96 (11.13) 76–117 
Memory California Verbal Learning Test (Delis, Kramer, Kaplan, & Ober, 1994) (CVLT-C)c   
Trial 1d,e −0.75 (0.84) −2–1 
Trial 1f 0.30 (0.76) −1–2 
Total score 46 (8.20) 2–59 
Children's Memory Scale (Cohen, 1997) (CMS)a   
Visual immediate 97 (19.16) 69–128 
Visual delayed 99 (13.96) 66–118 
Verbal immediate 92 (13.34) 63–115 
Verbal delayed 91 (11.42) 57–106 
General memory 92 (18.13) 59–118 
Learning 88 (16.55) 57–115 
Delayed recognition 97 (14.12) 69–118 
Rey Complex Figure (Waber & Holmes, 1986) (RCF)c   
Immediate recall 32 (8.41) 20–48 
Delayed Recall 31 (9.72) 20–48 
Recognitione 49 (7.85) 34–58 
Recognitionf 37 (8.98) 26–51 
Attention Conners’ Continuous Performance Test (Conners & Staff, 2000) (CPT)c   
Omissions 57 (15.06) 38–91 
Commissions 53 (6.73) 41–60 
Integrated Visual and Auditory Continuous Performance Test (IVA+Plus—Visual and Auditory Attention, n.d.) (IVA+Plus)a   
Full-scale response control quotient 76 (29.92) 0–110 
Auditory response control 82 (23.87) 18–106 
Visual response control 81 (27.66) 0–115 
Full-scale attention quotiente 57 (34.86) 0–102 
Full-scale attention quotientf 90 (5.38) 4–96 
Auditory attention 76 (24.17) 31–95 
Visual attention 71 (31.52) 0–113 
Domain Test Standard Score M (SDRange 
Intelligence Wechsler Intelligence Scale for Children IV (Wechsler, 2003) (WISC-IV): Compositesa   
Verbal Comprehension Index 94 (12.78) 67–121 
Perceptual Reasoning Index 90 (12.22) 65–112 
Working Memory Index 90 (15.02) 52–120 
Processing Speed Index 88 (15.85) 50–116 
Full-Scale IQ 87 (16.34) 50–115 
Wechsler Intelligence Scale for Children IV (WISC-IV): Subtestsb   
Similarities 10 (2.18) 5–13 
Vocabulary 9 (2.52) 3–13 
Comprehension 9 (2.63) 3–14 
Block design 7 (2.59) 3–14 
Picture concepts 9 (2.63) 2–13 
Matrix reasoning 9 (2.95) 3–16 
Digit span 8 (2.63) 2–13 
Letter number sequencing 9 (3.15) 1–14 
Coding 8 (3.13) 1–14 
Symbol search 8 (3.22) 1–13 
Academic achievement Woodcock–Johnson III Tests of Achievement (Woodcock, McGrew, & Mather, 2007) (WJ III ACH)a   
Math composite 91 (21.71) 45–124 
Reading composite 91 (22.34) 27–143 
Writing composite 97 (14.81) 72–130 
Letterword 100 (15.98) 79–148 
Passage Comprehension 92 (15.60) 65–138 
Calculation 93 (19.19) 57–124 
Spelling 97 (17.33) 68–130 
Writing sample 96 (11.13) 76–117 
Memory California Verbal Learning Test (Delis, Kramer, Kaplan, & Ober, 1994) (CVLT-C)c   
Trial 1d,e −0.75 (0.84) −2–1 
Trial 1f 0.30 (0.76) −1–2 
Total score 46 (8.20) 2–59 
Children's Memory Scale (Cohen, 1997) (CMS)a   
Visual immediate 97 (19.16) 69–128 
Visual delayed 99 (13.96) 66–118 
Verbal immediate 92 (13.34) 63–115 
Verbal delayed 91 (11.42) 57–106 
General memory 92 (18.13) 59–118 
Learning 88 (16.55) 57–115 
Delayed recognition 97 (14.12) 69–118 
Rey Complex Figure (Waber & Holmes, 1986) (RCF)c   
Immediate recall 32 (8.41) 20–48 
Delayed Recall 31 (9.72) 20–48 
Recognitione 49 (7.85) 34–58 
Recognitionf 37 (8.98) 26–51 
Attention Conners’ Continuous Performance Test (Conners & Staff, 2000) (CPT)c   
Omissions 57 (15.06) 38–91 
Commissions 53 (6.73) 41–60 
Integrated Visual and Auditory Continuous Performance Test (IVA+Plus—Visual and Auditory Attention, n.d.) (IVA+Plus)a   
Full-scale response control quotient 76 (29.92) 0–110 
Auditory response control 82 (23.87) 18–106 
Visual response control 81 (27.66) 0–115 
Full-scale attention quotiente 57 (34.86) 0–102 
Full-scale attention quotientf 90 (5.38) 4–96 
Auditory attention 76 (24.17) 31–95 
Visual attention 71 (31.52) 0–113 

Notes: The italicized tests are those for which we examined the impact of the various medical predictor variables.

aStandard Score M (SD) = 100 (15).

bStandard Score M (SD) = 10 (3).

cT-core M (SD) = 50 (10).

dZ-score.

eAcute lymphoblastic leukemia/ acute myeloid leukemia.

fSickle cell and other, nonasterisk tests report M and standard deviation of all disease categories.

Data Analysis

Data Consolidation

Preliminary data consolidation of variables was necessary. Diagnoses were grouped into categories to reflect similar therapy and risk factors encountered prior to transplant. The coding of diagnoses was as follows: (i) ALL (B cell, T cell, and acute leukemia of ambiguous lineage), (ii) myeloid leukemia (acute myeloid leukemia [AML] and juvenile myelomonocytic leukemia), (iii) sickle cell disease, and (iv) other nonmalignant diseases, which included severe congenital neutropenia, congenital amegakaryocytic thrombocytopenia, and congenital dyserythropoietic anemia type II. Stem cell source was coded as (i) related and unrelated bone marrow and peripheral stem cell donors and (ii) related and unrelated cord blood. Conditioning regimens were collapsed into (i) busulfan-based, non-TBI arm (busulfan/cytarabine/cyclophosphamide, busulfan/cyclophosphamide ± antithymocyte globulin) and (ii) TBI containing arm (cyclophosphamide/TBI, fludarabine/cyclophosphamide/TBI, and cyclophosphamide/thiotepa/TBI). aGVHD was coded dichotomously as (1) present or (0) absent. For severity analysis, Grade I patients were left out of the analysis secondary to low number of patients, with analysis focusing only on absent (Grade 0, coded as absent) and severe aGVHD (Grade II–IV, coded as severe). cGVHD was coded dichotomously as (1) present or (0) absent.

Neuropsychological tests were administered and scored in accordance with the respective manuals. Raw scores were converted to age- and gender-corrected standardized scores depending on the measure: Standard Scores (mean [M] = 100; standard deviation [SD] = 15); scaled scores (M = 10; SD = 3); T-scores (M = 50; SD = 10); Z-scores (M = 0; SD = 1). For the purposes of this study, the Reading Composite was based on an unweighted average of Letter-Word Identification, and Passage Comprehension and the Mathematics Composite was based on an unweighted average of the Calculation and Applied Problems subtests. Both composites were derived without the influence of the Fluency tests (Reading and Math Fluency). In this study, borderline impaired was defined as performance at or below 1.5 SDs below the mean and impaired was defined as performance at or below 2 SDs (Wefel, Vardy, Ahles, & Schagen, 2011).

Main Analyses

Descriptives

Descriptive statistics (frequencies, ranges, percentages, M, SD, and skew/kurtosis) were computed for all demographic, medical, and neuropsychological variables. As a preliminary step for the regression as well as the moderation and mediation analyses, Spearman correlation coefficients were computed to examine the relationships between the demographic, medical, and neuropsychological variables.

Group Level Analyses

In recognition that the assumptions of normal distribution would not be met with the small sample size, we used nonparametric tests and computed effect sizes for each analysis to indicate the magnitude of the mean differences between the comparison groups in SD units. Nonparametric Mann–Whitney U-tests or Kruskal–Wallis one-way analysis of variance or analysis of covariances were used to examine group differences in neuropsychological functioning as a function of demographic (e.g., age, gender, and time since transplant) and medical (e.g., diagnosis, type of transplant, source of transplant, conditioning regimen, grading, and severity of GVHD) variables based on the aforementioned data reduction methods. In addition, given that individuals with leukemia undergo both systemic and CNS-directed chemotherapy prior to conditioning therapy for HCT, we reexamined differences with the groups dichotomized (1 = ALL/AML and 2 = sickle cell and other). Although individuals with sickle cell may have underlying CNS vasculopathy, our small numbers prohibited separate analyses of this group. Our dichotomized groups did not differ significantly except on three variables: California Verbal Learning Test Trial 1, Rey Complex Figure Recognition, and Integrated Visual and Auditory (IVA) Attention Quotient. Consequently, for the analyses exploring neuropsychological performance as a function of demographic and medical predictor variables, we retained all groups, but conducted separate Mann–Whitney U tests for those three variables with the groups dichotomized. A Bonferroni correction was made due to multiple comparisons (α/24) and the critical p value established at .002 to minimize Type I error. However, given the small sample size, trends (p < .05) are also reported to minimize Type II error. Cohen's d effect sizes are also displayed to indicate the magnitude of effect. All subtests included in the analyses examining neuropsychological performance as a function of the medical variables are italicized in Table 1 (descriptives). In addition, given the retrospective nature of this study, not all participants completed all of these measures, hence the variable number of participants completing each test. For example, some participants may have completed the Children's Memory Scale to assess verbal and visual memory, whereas others would have completed the California Verbal Learning Test and Rey Complex Figure Test to assess similar domains.

Regression Analyses

Based on our a priori hypotheses and the bivariate relationships between the variables, we selected the Wechsler Intelligence Scale for Children-Fourth Edition (WISC-IV) Processing Speed and Children's Memory Scale Learning indices as vital predictor variables of the Mathematics and Reading Composite scores. The variables were entered into the equation using the forced entry method. Time-related variables such as age at transplant and time since transplant were not entered as covariates given that they were not significantly correlated with the outcome variables.

Regression Diagnostics

Regression diagnostics were conducted to ensure that the assumptions of regression were met and included an examination of (i) the variance of the predictors, (ii) multicollinearity diagnostics (Tolerance, Variance Inflation Factor, and Condition Index), (iii) the assumptions of linearity and homoscedasticity (plots of Z residuals against Z predicted values), (iv) independence of residuals (Durbin–Watson), (v) tests for normality of residuals (Normal P–P plots and histograms), and (vi) casewise diagnostics for undue influence (Cook's distance and Mahalanobis distance). The variance of the WISC-IV Processing Speed and Children's Memory Scale Learning Indices were nonzero. There was no evidence of significant multicollinearity between the predictor variables. The Variance Inflation Factor was 1.55, well below the suggested value of 10 (Bowerman & O'Connell, 1990). The Tolerance statistic was 0.65, whereas Tolerance below 0.2 suggests a potential problem with multicollinearity (Menard, 1995). The final collinearity diagnostics revealed well-distributed variance proportions of the predictors and a Condition Number of 1.19, whereas values of 30 are considered indicative of significant multicollinearity (Belsley, Kuh, & Welsch, 1980). An examination of the plots of Z residuals against Z predicted values revealed that the assumptions of linearity and homoscedasticity were met. The independence of residuals was examined with the Durbin–Watson statistic, which was 2.88 and 1.92 for the Mathematics and Reading models, respectively. Values between 1 and 3 suggest that the residuals in the models are independent. Tests for normality of residuals were examined with Normal P–P plots and histograms, which revealed generally normally distributed residuals. Casewise diagnostics for undue influence revealed no evidence of significant outliers or influential cases for either the Mathematics or Reading models (Cook's distance = 0.73 and 0.70 and Mahalanobis distance = 6.84 for both).

Mediation and Moderation Analyses

To test whether either GVHD or steroid treatment served as a mediator of the relationship between processing speed and/or memory with reading and/or mathematics performance, mediational analyses were conducted as described in Baron and Kenny (1986) and Hayes (2013). A series of regressions were computed for which we (i) regressed aGVHD or steroid treatment on academic functioning, (ii) regressed academic functioning on processing speed or memory, and (iii) regressed academic functioning on both the independent variables (processing speed or memory) and on the purported mediators (aGVHD and steroid treatment). We thought it was equally plausible that processing speed or memory served as a mediator of the relationship between GVHD or steroid treatment and academic outcome, so similar methods were used to examine this meditational model.

To test moderation, the impact of processing speed or memory as predictors (Path a), the impact of aGVHD or steroids as moderators (Path b), and the interaction or product of these two (Path c) was examined. All mediation and moderation analyses were conducted using the PROCESS procedure for SPSS Release 2.13 (Hayes, 2013).

Results

Participant Demographic and Medical Characteristics

A total of 30 participants were included in the study: 16 women and 14 men. Age at time of allogeneic transplant ranged between 1 and 17 years of age, with a mean age of 6. The age at time of neuropsychological assessment ranged from 4 to 17 years of age, with a mean age of 11. The time between transplant and assessment ranged from 1 to 14 years, with a mean interval of 5 years. The majority of participants were right-hand dominant (86%).

The majority of participants had leukemia (70%). An additional 10% had a diagnosis of sickle cell anemia, with the remaining 20% having a different nonmalignant diagnosis. Seventy percent of participants received related or unrelated allografts, with the remaining 30% receiving related or unrelated cord blood.

Prior to transplant, 53% of participants underwent a busulfan-based conditioning regimen, and 47% received a TBI-based regimen. Following transplantation, 50% of participants suffered from aGVHD and 47% from cGVHD. Of the overall sample, 37% and 40% received treatment with steroids for aGVHD and cGVHD, respectively. See Table 2 for a review of the demographic and medical characteristics of the sample.

Table 2.

Demographic and medical characteristics

Subject characteristic Frequency (nPercentage 
Gender 
 Men 16 53.3 
Donor type 
 Unrelated or related allografts 21 70.0 
 Unrelated or related cord blood 30.0 
Diagnosis 
 Lymphoid leukemiaa 14 46.7 
 Myeloid leukemiab 23.3 
 Sickle cell disease 10.0 
 Otherc 20.0 
Conditioning regimen 
 Busulfan based 16 53.3 
 Total body irradiation based 14 46.7 
Acute GVHD 
 None 15 50.0 
 Grade I 13.3 
 Grades II–IV 11 36.7 
Chronic GVHD 15 50.0 
 None 16 53.3 
 Present 14 46.7 
Steroids for acute GVHD 
 Yes 11 36.7 
 No 19 63.3 
Steroids for chronic GVHD 
 Yes 12 40.0 
 No 18 60.0 
 Range M (SD
Age 
 At HCT 1–17 6.07 (4.43) 
 At testing 4–20 11.07 (4.58) 
Time since HCT (years) 1–14 5.00 (4.05) 
Subject characteristic Frequency (nPercentage 
Gender 
 Men 16 53.3 
Donor type 
 Unrelated or related allografts 21 70.0 
 Unrelated or related cord blood 30.0 
Diagnosis 
 Lymphoid leukemiaa 14 46.7 
 Myeloid leukemiab 23.3 
 Sickle cell disease 10.0 
 Otherc 20.0 
Conditioning regimen 
 Busulfan based 16 53.3 
 Total body irradiation based 14 46.7 
Acute GVHD 
 None 15 50.0 
 Grade I 13.3 
 Grades II–IV 11 36.7 
Chronic GVHD 15 50.0 
 None 16 53.3 
 Present 14 46.7 
Steroids for acute GVHD 
 Yes 11 36.7 
 No 19 63.3 
Steroids for chronic GVHD 
 Yes 12 40.0 
 No 18 60.0 
 Range M (SD
Age 
 At HCT 1–17 6.07 (4.43) 
 At testing 4–20 11.07 (4.58) 
Time since HCT (years) 1–14 5.00 (4.05) 

Notes: GVHD = graft-versus-host disease; HCT = hematopoietic cell transplant.

aIncludes pre-B, pre-T, and acute leukemia of ambiguous lineage.

bIncludes acute myeloid leukemia and juvenile myelomonocytic leukemia.

cIncludes severe congenital neutropenia, congenital amegakaryocytic thrombocytopenia, and congenital dyserthropoietic anemia type II.

Descriptives of Neuropsychological Tests

The Mean Full-Scale IQ as measured by the WISC-IV was within 1 SD of the population mean, in the low average range. However, the Standard Scores ranged from 50 to 115, but 25% of the sample performed below the 10th percentile, suggesting that a substantive subset of children experienced significant impairment. Similarly, on the other indices, significant variability in our sample and substandard performance in a sizeable minority are apparent (see Table 1). Indeed, 17.4% of the sample performed at least 1 SD below average on verbal intellectual tasks, whereas 22% performed at or below the 10th percentile on perceptual-reasoning tasks. Similarly, on Working Memory and Processing Speed indices, 24% and 35%, respectively, also performed at or below the 10th percentile. These findings are consistent with our first hypothesis of inefficiencies in working memory and processing speed in a significant percentage of our sample.

Mean academic achievement abilities were all within the average range. However, consistent with intellectual test results, significant variability (Standard Score range: 45–143) within the composite scores was noted in our sample. In the areas of mathematics and reading, 26% and 22% scored at or below the 10th percentile, respectively. These data suggest substantial academic inefficiencies in a large number of our children consistent with our second hypothesis.

Mean performance on tests of sustained attention was variable and appeared test dependent. Although performance on one test of sustained visual attention (Continuous Performance Test) was within normal limits, scores on an alternate test that required both visual and auditory vigilance (IVA+Plus: Integrated Visual and Auditory Continuous Performance Test) ranged from low average to borderline. When the diagnostic groups were dichotomized, patients with ALL/AML displayed specific problems and grossly impaired performance on the Full Scale Attention Quotient.

On a test of visual memory (Rey Complex Figure Test), performance was impaired and ∼2 SDs below the mean for both immediate and delayed recall. On a recognition trial, those with ALL/AML generally performed within the average range, whereas those with sickle cell disease and other nonmalignant disorders performed in the borderline to impaired range. This suggests that the former may have more specific problems with retrieval rather than encoding mechanisms. On additional measures of verbal and visual memory functioning (California Verbal Learning Test and Children's Memory Scale), performance was grossly within the average range. Nonetheless, again significant variability was apparent on these measures. Of note, a full third of the sample performed at or below the 10th percentile on the General Memory and Learning Indices of the CMS. These findings are consistent with our first hypothesis.

Group Level Analyses

Effects of Time-based Variables

Time-based variables that might have affected neuropsychological functioning were age at HCT, interval between HCT and neuropsychological testing, and year transplanted (as an indicator of possible cohort effects related to changes in treatment over time). Due to sample size constraints, bivariate correlations rather than multivariate methods were used.

Effects of Age at HCT

Intelligence and achievement. Age at HCT was not related to composite indices of intelligence, although older children performed significantly better on the Picture Concepts subtest (r = .46, p = .03). Age at HCT did not affect any tests of academic achievement.

Learning, memory, and attention. No effects of age at HCT were apparent.

Effects of Length of Interval Between HCT and Neuropsychological Testing

Intelligence and achievement. Time since HCT was negatively correlated with the Perceptual Reasoning Index (r =− .41, p < .05) such that the longer the interval, the worse the performance. There were no significant relationships between time since HCT and the composite indices of Verbal Comprehension, Working Memory, Processing Speed, or the Full-Scale IQ. Time since HCT was also negatively correlated with the Comprehension subtest (r =− .43, p < .05). Time since HCT did not affect most tests of achievement, although a significant negative relationship was evident on a test of applied math suggesting poorer performance was evident over time (r =− .83, p < .05).

Learning, memory, and attention. Time since transplant was negatively correlated with the Children's Memory Scale measure of delayed recall (r =− .45, p < .05) such that the longer the time since transplant, the poorer the performance.

Effects of Year of HCT

Year of transplant positively correlated with the Composite Index of Working Memory (r = .48, p < .05) and three domains of the Children's Memory Scale (Verbal Immediate, r = .47, p < .05; Delayed Recall, r = .55, p < .01; and General Memory, r = .59, p ≤ .01). In each case, more recent transplant was correlated with better performance. This relative resilience of verbal skills may be indicative of changes in treatment regimen over time. As expected, year of transplant was not correlated with age at HCT. However, not unexpectedly, as current clinical practice has evolved and more patients are referred earlier for neuropsychological testing, time elapsed between transplant and assessment strongly correlated with year of transplant (r =− .88, p < .001).

Neuropsychological Results as a Function of Medical Variables

Reported subsequently are the neurocognitive domains exhibiting differential effects as a function of medical variables. Performance across all neuropsychological domains as a function of medical variables is displayed in Table 3. Effect sizes are large (Cohen, 1988), ranging from 0.92 to 1.64.

Table 3.

Performance across neuropsychological domain by medical variable

Medical variable Neuropsychological test n M (SDM (SDU p d 
Acute GVHD   aGVHD No aGVHD    
Memory 
CVLT trial 1a 16 −0.21 (0.64) −1.17 (0.75) 52.00 .031* 1.38 
CMS verbal delayed 21 96 (8.12) 87 (12.64) 92 .040* 0.92 
Acute GVHD severity level    Grades 0–I  Grades II–IV    
Memory 
CVLT trial 1a 16 −1.15 (71) −0.08 (0.59) 52.50 .011* 1.64 
CMS general memory 17 84 (14.34) 102 (16.56) 74 .017* 1.18 
CMS delayed recognition 19 91 (13.95) 106 (9.54) 99 .006** 1.30 
Intellectual-WISC-4 
Perceptual Reasoning Index 20 87 (11.50) 98 (8.71) 97 .033* 1.02 
    Steroids  No steroids    
 Memory 
Steroids for acute GVHD CMS general memory 18 102 (18.96) 85 (14.94) 67 .036* 1.06 
CMS delayed recognition 21 107 (9.12) 92 (13.72) 87 .014* 1.06 
Medical variable Neuropsychological test n M (SDM (SDU p d 
Acute GVHD   aGVHD No aGVHD    
Memory 
CVLT trial 1a 16 −0.21 (0.64) −1.17 (0.75) 52.00 .031* 1.38 
CMS verbal delayed 21 96 (8.12) 87 (12.64) 92 .040* 0.92 
Acute GVHD severity level    Grades 0–I  Grades II–IV    
Memory 
CVLT trial 1a 16 −1.15 (71) −0.08 (0.59) 52.50 .011* 1.64 
CMS general memory 17 84 (14.34) 102 (16.56) 74 .017* 1.18 
CMS delayed recognition 19 91 (13.95) 106 (9.54) 99 .006** 1.30 
Intellectual-WISC-4 
Perceptual Reasoning Index 20 87 (11.50) 98 (8.71) 97 .033* 1.02 
    Steroids  No steroids    
 Memory 
Steroids for acute GVHD CMS general memory 18 102 (18.96) 85 (14.94) 67 .036* 1.06 
CMS delayed recognition 21 107 (9.12) 92 (13.72) 87 .014* 1.06 

Notes: GVHD = graft-versus-host disease; CVLT = California Verbal Learning Test; CMS = Children's Memory Scale; WISC-4 = Wechsler Intelligence Scale for Children-Fourth Edition.

aGroup differences reevaluated only including AML/ALL diagnoses.

*p < .05; **p < .01.

aGVHD Severity

Unexpectedly, patients with aGVHD Grades II–IV displayed better performance than those without aGVHD on the Children's Memory Scale Delayed Recognition Index (U(1,18) = 99, p = .006, d = 1.30) and showed trends in the same direction on the Children's Memory Scale General Memory Index (U(1,16) = 74, p = .02, d = 1.18) and California Verbal Learning Test Trial 1 (U(1,15) = 52.5, p = .01, d = 1.64).

Steroids

Participants who received steroidal treatment for aGVHD trended toward better performance on the Children's Memory Scale Delayed Recognition Index (U(1,19) = 87, p = .01, d = 1.06) than those who received no steroids for aGVHD. Nonetheless, the latter group performance was at the low end of the average range.

Diagnosis, Donor Type, Conditioning, and GVHD

There were no significant differences on any of the neuropsychological measures as a function of diagnosis (ALL vs. AML). Likewise, there were no significant differences in intellect, achievement, attention, or memory as a function of donor type (cord blood transplant vs. allografts), conditioning (busulfan-based therapy vs. TBI), or incidence of GVHD (acute or chronic).

Regression Analyses Predicting Academic Performance

Regression Analyses

Results of the regression analyses revealed that the Processing Speed Index and Children's Memory Scale Learning Index were significant predictors of the Mathematics Composite, accounting for 69% of the variance (p < .001). In contrast, only the Children's Memory Scale Learning Index was a significant predictor of the Reading Composite in a model accounting for 61% of the variance (p = .01). These results were consistent with our third hypothesis that the core process variables of processing speed and memory would account for a significant amount of the variance in macrolevel academic outcomes. See Table 4 for complete results.

Table 4.

Linear multiple regression analysis predicting academic achievement with processing speed and memory

 B SE B R2 β p 
Mathematics model   .69  <.001 
 Constant −4.53 16.48    
 WISC-IV PSI 0.53 0.19  0.47 .02 
 CMS Learn 0.55 0.21  0.46 .02 
Reading model   .61  .001 
 Constant 21.02 15.24    
 WISC-IV PSI 0.32 0.18  0.36 .09 
 CMS Learn 0.51 0.19  0.52 .02 
 B SE B R2 β p 
Mathematics model   .69  <.001 
 Constant −4.53 16.48    
 WISC-IV PSI 0.53 0.19  0.47 .02 
 CMS Learn 0.55 0.21  0.46 .02 
Reading model   .61  .001 
 Constant 21.02 15.24    
 WISC-IV PSI 0.32 0.18  0.36 .09 
 CMS Learn 0.51 0.19  0.52 .02 

Note: CMS Learn = Children's Memory Scale Learning Index; SE B = Standard Error Beta; WISC-IV PSI = Wechsler Intelligence Scale for Children-Fourth Edition, Processing Speed Index.

Mediation and Moderation Analyses

We examined both aGVHD severity and steroidal treatment as well as speed of information processing and memory as possible mediators or moderators of reading and math performance, with consistently nonsignificant findings.

Discussion

Given the increasing population of transplant survivors and the high prevalence of morbidity despite cure of the original disease, a great deal of effort is now being devoted to characterizing and preventing late complications, including those affecting cognitive and psychosocial functioning. Child HCT survivors are at increased risk of chronic medical problems due to a number of variables, including intensity of chemotherapy, TBI, inflammatory processes, and toxicities related to immune suppression. These same variables also place children who undergo HCT at risk for long-term neurocognitive sequelae. Prior investigations examining late neurocognitive effects of children following HCT have focused on intellectual functioning and more limited aspects of neuropsychological functioning over relatively brief post-treatment periods, with limited consideration of individual variability. Our study sought to illuminate the individual variability in neurocognitive functioning and more precisely identify predictors of academic outcomes across a relatively longer post-HCT trajectory. Our medical variables were well characterized so that we could examine relationships and differences in cognitive outcome based on potential medical predictors of neurotoxicity. We further sought to determine if the core process variables of processing speed and memory could account for a substantial percentage of the variance in outcome in academic performance in children post-HCT.

The demographics of the pediatric population examined in this investigation are consistent with prior studies. On average, children were ∼6 years of age at the time of HCT and underwent a comprehensive evaluation ∼5 years post-HCT.

Overall, as a group, intellect was within the average range, with a relative weakness noted in processing speed. Mean intellect and academic achievement hovered in the average to low average ranges, masking the rather profound individual variability to the extent that approximately one fourth of the sample performed in the borderline impaired range and below. This variability suggests long-term risk and resilience factors in children undergoing HCT, which are still poorly understood. Similarly, although mean verbal and visual learning were in the average to low average ranges, up to one third of the sample performed in the borderline impaired range and below. In addition, children performed significantly more poorly on a visual memory task (Rey Complex Figure Test) that also required a greater degree of executive skills such as planning and organizing. This may reflect greater vulnerability when higher cognitive load is required. Similarly, performance was borderline to impaired on a test for which task demands required relatively complex attentional shifting between auditory and visual stimuli (IVA+Plus: Integrated Visual and Auditory Continuous Performance Test).

Age at time of transplant was unrelated to most neurocognitive variables, consistent with Phipps and colleagues' (Phipps, Rai, Leung, Lensing, & Dunavant, 2008) report of no significant age effect on outcome in children through 5 years post-stem cell transplantation.

Time since transplant was related to increased attentional difficulties. Consistent with some prior prospective investigations (Kramer et al., 1997; Shah et al., 2008), we also found time (since transplant) associated with statistically poorer performance on selected tests of intellectual functioning, specifically perceptual-reasoning. Kramer and colleagues (1997) reported a decrement of >2 SDs from baseline at 1-year post-HCT and no further deterioration at 3 years post-HCT. In contrast, a substantial subset of our sample manifested borderline intellectual impairment or worse performance at a mean of ∼5 years post-HCT.

Academic vulnerability has been previously reported in studies examining long-term outcome following HCT (Barrera & Atenafu, 2008) and in other pediatric populations following disruptions from a host of CNS disorders (Collins & Rourke, 2003). In our study, time since transplant was unrelated to reading or written language skills at a mean of ∼5 years post-HCT, although a trend to poorer performance was evident on tests of applied mathematics. Additionally, a greater degree of general memory impairment and poorer delayed recall were evident over time as well, consistent with the decline in memory functioning over time reported by Shah and colleagues (2008) at a 5-year follow-up.

Notably, the current study contributes to the literature by specifically examining the extent to which the core neurocognitive processes of information processing speed and memory affect more global academic outcome in children who have undergone HCT. Our finding that 69% of the variance in mathematics is accounted for by processing speed and memory performance, whereas 61% of the variance in reading is accounted for by memory is novel to the HCT literature. This mirrors the broader outcome literature in children with CNS disruption from a number of etiologies, neurodevelopmental or acquired. In summary, our research and that of others tentatively implies that a subset of pediatric HCT survivors experience a memory deficit or inefficiencies in memory functioning, which may possibly decrease their ability to acquire new information. Whether these apparent relationships hold up could be better investigated in studies employing prospective longitudinal designs and causal structural equation modeling or path analysis.

Our findings indicate no differences in intellectual functioning in relation to the medical predictors. First, with regard to type of transplant, although there were no statistically significant differences between children who underwent transplantation via bone marrow or peripheral blood versus cord blood donors, a further examination of their intellectual profiles revealed that the former performed ∼1 SD below the standardization mean compared with cord blood recipients, who performed solidly within the average range.

With regard to conditioning regimens, children in our study who underwent a busulfan-based conditioning regimen did not differ from those receiving TBI on tasks of attention and memory. Our findings contrast with those of Levy and colleagues (2013). In their retrospective review of outcomes of children <3 years of age, it was found that those who received TBI as part of their preparative regimen for HCT had more significant multiorgan dysfunction, including cognitive impairment and language and/or speech delay, at 1-year follow-up. Clearly, the effect of conditioning therapy and underlying diagnoses on neurocognitive outcomes requires future investigation with a larger sample.

Consistent with other investigations examining potential predictors of outcome following HCT, the severity of GVHD appears to be a crucial, yet complex variable, and its relationship to cognitive outcome and neurotoxicity are poorly understood in pediatrics. Our study revealed that the severity of GVHD generally appeared to affect memory functioning; however, our findings appear to be paradoxical. In our retrospective cohort, children who had experienced aGVHD at a greater level of severity performed significantly better on tests of initial verbal learning, general memory, and delayed recall than those without aGVHD. These findings would appear to contrast with Phipps and colleagues' (2008) report of substantial declines in intellectual abilities in transplant recipients with aGVHD over time. Interestingly, children in our sample who underwent steroid treatment for aGVHD also demonstrated significantly better delayed recognition performance. In general, those who experienced aGVHD with a more severe presentation and who underwent steroidal treatment performed ∼1 SD better on tests of memory functioning up to 5 years post-transplant. Through exploratory analyses, we examined both aGVHD severity and steroidal treatment as well as speed of information processing and memory as possible mediators or moderators of reading and math performance. The consistently nonsignificant findings suggest that elucidating the source of the additional variance in these outcomes will require the examination of more complex mediation and/or moderation models, which would only be possible with a considerably larger sample size.

To our knowledge, this is the first time that differential memory enhancement has been reported in children with versus without aGVHD. It is well established that the immune system plays an important role in normal and pathologic brain processes. Immune-mediated brain processes include regulation of non-neuronal cells, modulation of synaptic strength, regulation of neurotransmission, and direct involvement in cellular processes leading to neuronal apoptosis. Under normal conditions, the immune system provides a stimulatory environment to promote neuronal plasticity and learning, consolidation of memories, and hippocampal long-term potentiation. However, alterations in the immune system including T-cell depletion and abnormal cytokine levels, two processes that are inherent in the HCT procedure, can be detrimental to neurocognitive processes. As patients slowly engraft the donor immune system, there may be variable periods of increased susceptibility to neurotoxicity as both complete absence and overproduction of cytokines can lead to poor neurologic functioning. Development of aGVHD and steroid treatment may further disrupt the delicate balance of T-cell number, function, and cytokine production critical for hippocampal function and memory consolidation. These factors may contribute to the rather substantial variability in memory performance in our sample, but the complex impact of GVHD and immunosuppressant therapy on neurocognition will require more rigorous examination in future studies.

This study has a number of limitations. One may be that only those children experiencing neurocognitive deficits were referred for testing, thus suggesting an acquisition bias in our retrospective review. Our sample size was also small, thus limiting external validity, and we did not include a medical comparison or control population. Conducting large prospective pediatric studies in HCT survivors is difficult given the rarity of patients requiring HCT as well as the high morbidity and mortality rates post-HCT that result in attrition of available subjects. Consider that for patients with AML or ALL undergoing HCT, the 3-year probabilities for survival range from 67% for low-risk disease to only 26% for high-risk disease (Pasquini, Wang, Howowitz, & Gale, 2013). Nevertheless, our sample size and study design is comparable with other investigations of long-term neurocognitive outcome in children who have undergone HCT (Perkins et al., 2007; Shah et al., 2008) who have also focused on better characterization of this population without a clinical control. Complicating our small sample size is the relatively large heterogeneity of our sample in regard to diagnosis, limiting meaningful comparisons between disease groups. In addition, more significant group differences in medical and demographic predictors such as maternal age or socioeconomic status may have been apparent in a prospectively designed investigation. Future longitudinal research with larger samples would ideally control for demographic variables and investigate the possible moderation of cognitive outcomes by strategically selected medical variables. We are currently investigating neurocognitive outcomes in a prospective trial, which will help overcome some of the limitations of this study.

In summary, the intent of this investigation was to better clarify and characterize neurocognitive predictors of academic outcome in children who had undergone HCT and were referred for comprehensive neuropsychological testing. Differences observed in neurocognitive functioning as a function of age at transplant and time since transplant were in the expected direction, with poorer outcome noted over time, but with age at transplant appearing to be a less important predictor. We offer some preliminary findings on the effect of type and source of transplantation and conditioning regimen on cognition, but further investigation is necessary. Our results reveal important neurocognitive variables, specifically processing speed and memory, which may strongly influence academic outcomes in children following HCT. This has very practical and important implications for the successful reintegration of children following HCT. Younger children will clearly benefit from additional monitoring and early intervention to minimize potential negative long-term consequences of their illness and treatment. School-aged children will likely benefit from strategies to compensate for memory and processing speed inefficiencies, such as using notes or securing extended time for academic assignments. Mathematics performance should be closely monitored for possible learning disability, so timely services can also be secured. Finally, our investigation may illuminate a potential resilience or neurotrophic factor for memory as not all patients demonstrated defects on neuropsychological testing. Further investigation into these nuances is crucial as memory is vital to all aspects of learning (Moscovitch, Chein, Talmi, & Cohn, 2007) and exploring both remedial and compensatory efforts to enhance memory skills essential for these pediatric HCT survivors.

Conflict of Interest

None declared.

References

Abou-Mourad
Y. R.
,
Lau
B. C.
,
Barnett
M. J.
,
Forrest
D. L.
,
Hogge
D. E.
,
Nantel
S. H.
et al
. (
2010
).
Long-term outcome after allo-SCT: Close follow-up on a large cohort treated with myeloablative regimens
.
Bone Marrow Transplantation
 ,
45
,
295
302
.
Anderson
V.
,
Godfrey
C.
,
Rosenfeld
J. V.
,
Catroppa
C.
(
2012
).
Predictors of cognitive function and recovery 10 years after traumatic brain injury in young children
.
Pediatrics
 ,
129
,
e254
e261
.
Armenian
S. H.
,
Sun
C.-L.
,
Kawashima
T.
,
Arora
M.
,
Leisenring
W.
,
Sklar
C. A.
et al
. (
2011
).
Long-term health-related outcomes in survivors of childhood cancer treated with HSCT versus conventional therapy: A report from the Bone Marrow Transplant Survivor Study (BMTSS) and Childhood Cancer Survivor Study (CCSS)
.
Blood
 ,
118
,
1413
1420
.
Armstrong
D. F.
(
2001
).
Acute and long-term neurodevelopmental outcomes in children following bone marrow transplantation
.
Frontiers in Bioscience
 ,
6
,
G6
G12
.
Arnett
A. B.
,
Peterson
R. L.
,
Kirkwood
M. W.
,
Taylor
H. G.
,
Stancin
T.
,
Brown
T. M.
et al
. (
2013
).
Behavioral and cognitive predictors of educational outcomes in pediatric traumatic brain injury
.
Journal of the International Neuropsychological Society
 ,
19
,
881
889
.
Baron
R. M.
,
Kenny
D. A.
(
1986
).
Moderator-mediator variable distinction in social psychological research: Conceptual, strategic, and statistic considerations
.
Journal of Personality and Social Psychology
 ,
51
,
1173
1182
.
Barrera
M.
,
Atenafu
E.
(
2008
).
Cognitive, educational, psychosocial adjustment and quality of life of children who survive hematopoietic SCT and their siblings
.
Bone Marrow Transplantation
 ,
42
,
15
21
.
Bejanyan
N.
,
Weisdorf
D. J.
,
Logan
B. R.
,
Wang
H.-L.
,
Devine
S. M.
,
de Lima
M.
et al
. (
2015
).
Survival of patients with acute myeloid leukemia relapsing after allogeneic hematopoietic cell transplantation: A Center for International Blood and Marrow Transplant Research Study
.
Biology of Blood and Marrow Transplantation
 ,
21
,
454
459
.
Belsley
D. A.
,
Kuh
E.
,
Welsch
R. E.
(
1980
).
Regression diagnostics: Identifying influential data and sources of collinearity
 .
New York
:
John Wiley and Sons
.
Bowerman
B. L.
,
O'Connell
R. T.
(
1990
).
Linear statistical models: An applied approach
  (
2nd ed.
).
Belmont, CA
:
Duxbury
.
Cohen
J.
(
1988
).
Statistical power analysis for the behavioral sciences
  (
2nd ed.
).
Hillsdale, NJ
:
Erlbaum
.
Cohen
M.
(
1997
).
Childrens memory scale
 .
San Antonio, TX
:
The Psychological Corporation
.
Collins
D.
,
Rourke
B.
(
2003
).
Learning-disabled brains: A review of the literature
.
Neuropsychology, Development, and Cognition. Section A, Journal of Clinical and Experimental Neuropsychology
 ,
25
,
1011
1034
.
Conners
C. K.
,
Staff
M.
(Eds.). (
2000
).
Conners’ Continuous Performance Test II: Computer program for windows technical guide and software manual
 .
North Tonawanda, NY
:
Multi-Health System
.
Delis
D. C.
,
Kramer
J. H.
,
Kaplan
E.
,
Ober
A.
(
1994
).
The California Verbal Learning Test-Children’s Version
 .
San Antonio, TX
:
The Psychological Corporation
.
Ferry
C.
,
Gemayel
G.
,
Rocha
V.
,
Labopin
M.
,
Esperou
H.
,
Robin
M.
et al
. (
2007
).
Long-term outcomes after allogeneic stem cell transplantation for children with hematological malignancies
.
Bone Marrow Transplantation
 ,
40
,
219
224
.
Hayes
A. F.
(
2013
).
Introduction to mediation, moderation, and conditional process analysis: A regression-based approach
 .
New York
:
Guilford Publications
.
IVA+Plus—Visual and Auditory Attention
 . (
n.d.
).
North Chesterfield, VA: Brain Train Incorporated
.
Kodituwakku
P. W.
(
2007
).
Defining the behavioral phenotype in children with fetal alcohol spectrum disorders: A review
.
Neuroscience and Biobehavioral Reviews
 ,
31
,
192
201
.
Kramer
J. H.
,
Crittenden
M. R.
,
DeSantes
K.
,
Cowan
M. J.
(
1997
).
Cognitive and adaptive behavior 1 and 3 years following bone marrow transplantation
.
Bone Marrow Transplantation
 ,
19
,
607
613
.
Kupst
M. J.
,
Penati
B.
,
Debban
B.
,
Camitta
B.
,
Pietryga
D.
,
Margolis
D.
et al
. (
2002
).
Cognitive and psychosocial functioning of pediatric hematopoietic stem cell transplant patients: A prospective longitudinal study
.
Bone Marrow Transplantation
 ,
30
,
609
617
.
Levy
J. M. M.
,
Tello
T.
,
Giller
R.
,
Wilkening
G.
,
Quinones
R.
,
Keating
A. K.
et al
. (
2013
).
Late effects of total body irradiation and hematopoietic stem cell transplant in children under 3 years of age
.
Pediatric Blood & Cancer
 ,
60
,
700
704
.
Menard
S.
(
1995
).
Applied logistic regression analysis. Sage university paper series on quantitative applications in the social sciences, 07–106
 ,
Thousand Oaks, CA: Sage
.
Moscovitch
M.
,
Chein
J. M.
,
Talmi
D.
,
Cohn
M.
(
2007
).
Learning and memory
. In
Baars
B. J.
,
Gage
N. M.
(Eds.),
Cognition, brain and consciousness
 .
London
:
Elsevier
.
Murray
A. L.
,
Scratch
S. E.
,
Thompson
D. K.
,
Inder
T. E.
,
Doyle
L. W.
(
2014
).
Neonatal brain pathology predicts adverse attention and processing speed outcomes in very preterm and/or very low birth weight children
.
Neuropsychology
 ,
28
,
552
562
.
Pasquini
M.
,
Wang
Z.
,
Howowitz
M. M.
,
Gale
R. P.
(
2013
).
2013 report from the Center for International Blood and Marrow Transplant Research (CIBMTR): Current uses and outcomes of hematopoietic cell transplants for blood and bone marrow disorders
.
Clinical Transplant
 ,
2013
,
187
197
.
Perkins
J. L.
,
Kunin-Batson
A. S.
,
Youngren
N. M.
,
Ness
K. K.
,
Ulrich
K. J.
,
Hansen
M. J.
et al
. (
2007
).
Long-term follow-up of children who underwent hematopoietic cell transplant (HCT) for AML or ALL at less than 3 years of age
.
Pediatric Blood & Cancer
 ,
49
,
958
963
.
Peterson
C. C.
,
Johnson
C. E.
,
Ramirez
L. Y.
,
Huestis
S.
,
Pai
A. L. H.
,
Demaree
H. A.
et al
. (
2008
).
A meta-analysis of the neuropsychological sequelae of chemotherapy-only treatment for pediatric acute lymphoblastic leukemia
.
Pediatric Blood & Cancer
 ,
51
,
99
104
.
Phipps
S.
,
Rai
S. N.
,
Leung
W.-H.
,
Lensing
S.
,
Dunavant
M.
(
2008
).
Cognitive and academic consequences of stem-cell transplantation in children
.
Journal of Clinical Oncology: Official Journal of the American Society of Clinical Oncology
 ,
26
,
2027
2033
.
Pingali
S.
,
Champlin
R.
(
2015
).
Pushing the envelope—nonmyeloablative and reduced intensity preparative regimens for allogeneic hematopoietic transplantation
.
Bone Marrow Transplantation
 ,
50
,
1157
1167
.
Rasche
L.
,
Kapp
M.
,
Einsele
H.
,
Mielke
S.
(
2014
).
EBV-induced post transplant lymphoproliferative disorders: A persisting challenge in allogeneic hematopoetic SCT
.
Bone Marrow Transplantation
 ,
49
,
163
167
.
Schatz
J.
,
Kramer
J. H.
,
Ablin
A.
,
Matthay
K. K.
(
2000
).
Processing speed, working memory, and IQ: A developmental model of cognitive deficits following cranial radiation therapy
.
Neuropsychology
 ,
14
,
189
200
. Retrieved from .
Shah
A. J.
,
Epport
K.
,
Azen
C.
,
Killen
R.
,
Wilson
K.
,
De Clerck
D.
et al
. (
2008
).
Progressive declines in neurocognitive function among survivors of hematopoietic stem cell transplantation for pediatric hematologic malignancies
.
Journal of Pediatric Hematology/Oncology
 ,
30
,
411
418
.
Simms
S.
,
Kazak
A. E.
,
Golomb
V.
,
Goldwein
J.
,
Bunin
N.
(
2002
).
Cognitive, behavioral, and social outcome in survivors of childhood stem cell transplantation
.
Journal of Pediatric Hematology/Oncology
 ,
24
,
115
119
. Retrieved from .
Sun
C.-L.
,
Francisco
L.
,
Kawashima
T.
,
Leisenring
W.
,
Robison
L. L.
,
Baker
K. S.
et al
. (
2010
).
Prevalence and predictors of chronic health conditions after hematopoietic cell transplantation: A report from the Bone Marrow Transplant Survivor Study
.
Blood
 ,
116
,
3129
3139
;
quiz 3377
.
Waber
D. P.
,
Holmes
J. M.
(
1986
).
Assessing children's memory productions of the Rey-Osterrieth Complex Figure
.
Journal of Clinical and Experimental Neuropsychology
 ,
8
,
563
580
.
Wechsler
D.
(
2003
).
Wechsler Intelligence Scale for Children, Fourth Edition (WISC-IV)
 .
San Antonio, TX: Pearson Assessment
.
Wefel
J. S.
,
Vardy
J.
,
Ahles
T.
,
Schagen
S. B.
(
2011
).
International cognition and Cancer Task Force recommendations to harmonise studies of cognitive function in patients with cancer
.
The Lancet Oncology
 ,
12
,
703
708
.
Woodcock
R. W.
,
McGrew
K. S.
,
Mather
N.
(
2007
).
Woodcock Johnson Tests of Achievement-III
 .
Rolling Meadows, IL
:
Riverside Publishing
.
Worel
N.
,
Biener
D.
,
Kalhs
P.
,
Mitterbauer
M.
,
Keil
F.
,
Schulenburg
A.
et al
. (
2002
).
Long-term outcome and quality of life of patients who are alive and in complete remission more than two years after allogeneic and syngeneic stem cell transplantation
.
Bone Marrow Transplantation
 ,
30
,
619
626
.

Author notes

Co-first authors.
Senior author.