Characterization of Leukoencephalopathy and Association with Later Neurocognitive Performance in Pediatric Acute Lymphoblastic Leukemia

Abstract

Objectives:

The most common form of pediatric cancer is acute lymphoblastic leukemia (ALL). Magnetic resonance (MR) neuroimaging studies have revealed leukoencephalopathy (LE) in pediatric ALL, but the impact of LE on long-term neurocognitive performance remains unknown. This study aims to objectively characterize the prevalence, extent, and intensity of LE, and their association with later neurocognitive performance.

Materials and Methods:

Pediatric patients (N=377) treated for ALL without irradiation underwent MR neuroimaging at four time points throughout therapy (end of remission induction (MR1), end of consolidation (MR2), week 31 (MR3) and week 120 (end therapy; MR4) of continuation treatment and neurocognitive evaluations at the end of therapy and two years later. Generalized Estimation Equation (GEE) models with logit link were developed to explore the association between LE prevalence and extent with time points throughout therapy, age at diagnosis (≤5 years or >5 years), treatment risk arm (low (LR) or standard/high (SHR) risk), and sex. General linear models were also developed to investigate the association between neuroimaging metrics during treatment and neurocognitive performance at two-year follow-up.

Results:

Prevalence of LE was greatest (22.8%; 74/324) following consolidation therapy. Prevalence of LE increased at MR2 relative to MR1 regardless of treatment risk arm (both: p<0.001), age group (both: p<0.001), or sex (males: p<0.001, females: p=0.013). Extent of white matter affected also increased at MR2 relative to MR1 regardless of treatment risk arm (SHR: p<0.001, LR: p=0.004), age group (both: p<0.001), or sex (male: p<0.001, female p=0.001). Quantitative relaxation rates were significantly longer in LE compared to normal-appearing white matter in the same examination (T1: p<0.001; T2: p<0.001). LE prevalence early in therapy was associated with increased parent ratings of conduct problems (p=0.039) and learning difficulties (p=0.036) at two-year follow-up compared with end of therapy. Greater extent of LE early in therapy was associated with decreasing performance on a measure of processing speed (p=0.003) from end of therapy to two-year follow-up. A larger extent of LE at the end of therapy was associated with decreased performance in reading (p=0.004), spelling (p=0.003) and mathematics (p=0.019) at two-year follow-up and increasing problems with attention (Omissions (p=0.045); Beta (p=0.015)) and memory (List A Total Recall (p=0.010)) at two year follow up compared with end of therapy.

Conclusions:

In this large cohort of pediatric patients treated for ALL without irradiation, asymptomatic LE during therapy can be seen in almost a quarter of patients, involve as much as 10% of the WM volume, and is associated with decreasing neurocognitive performance, increasing parent reports of conduct problems and learning difficulties in survivors.

1. Introduction

Acute lymphoblastic leukemia (ALL) is the most common pediatric cancer and accounts for more than 3,000 new cases each year in the US (1). Methotrexate (MTX) is an essential chemotherapeutic drug, which is frequently administered both intrathecally and at high doses (HDMTX) intravenously as consolidation treatment, and contributes to contemporary cure rates approaching 90% in pediatric ALL (2). However, MTX can cause clinical neurotoxicity by disrupting CNS folate homeostasis and/or direct neuronal damage, impeding the process of myelination and, subsequently, neuronal transmission (3–6). Magnetic resonance (MR) neuroimaging studies reveal leukoencephalopathy (LE) to be the most common methotrexate-related neurotoxic side effect and has been associated with increased MTX exposure (7, 8).

LE is seen as T2-weighted white matter (WM) hyper-intensities and may be transient or chronic (9–11). MR image contrast reflects differences in T1 and T2 relaxation times in LE, resulting from a variety of biological underpinnings (12). Previous studies have shown increases in T1 and T2 relaxation rates in LE compared to normal-appearing white matter (NAWM), which were significantly correlated with each other and were dependent on the proportion of WM affected (13, 14).

Neurocognitive deficits in ALL patients can have a devastating effect on long-term quality of life (15). Most commonly reported sequelae are attention deficits and working memory (16–19). Longer-term disruptions in neurocognitive functioning can lead to lower academic performance and decreased social attainment, which are associated with clinical and demographic factors (15, 20, 21).

A recent study identified clinical, pharmacokinetic and genetic risk factors for treatment-induced neurotoxicity (22) in a large cohort of children treated for ALL at a single institution on a uniform protocol-directed, risk-stratified treatment regimen that did not include irradiation. Additionally, higher doses of MTX and younger age at diagnosis have previously been associated with decreased neurocognitive performance at the end of therapy (15) and at a two-year follow-up (23) in this same cohort of patients. However, the impact of early imaging changes on long-term neurocognitive performance remains to be evaluated. The purpose of this study was to first characterize the prevalence, extent, and intensity of LE during therapy and then investigate the hypothesis that these early imaging markers will be associated with later neurocognitive performance.

2. Materials and Methods

2.A. Patients and Treatment Protocol

Between June 2000 and October 2007, 498 pediatric ALL patients were enrolled on an institutional frontline treatment protocol (NCT00137111) and were approached for participation in this prospective, longitudinal MRI screening study. The study was approved by the Institutional Review Board and conducted in accordance with guidelines from the National Cancer Institute and Office for Human Research Protection. Research imaging was integrated into protocol-directed clinical care for all patients at least one year of age at diagnosis. Informed consent was obtained from each patient’s parent or guardian. Among these patients, 391 volunteered to participate. Fourteen patients were excluded: ten based on a Down syndrome diagnosis, and one each due to early withdrawal from the protocol, cerebral thrombosis, brain cyst, and brain hematoma. Data from 377 (207 males) children, ages 1–18 years (mean=6.9±4.6) were analyzed.

The institutional treatment protocol has previously been described (24) and is demonstrated graphically in Figure 1. All patients received triple IT chemotherapy with MTX, cytarabine, and hydrocortisone as CNS-directed therapy beginning with remission induction. HDMTX was given as one course at 1.0 g/m² on day 1 of remission induction, and four additional courses given upon achieving completion of remission induction as consolidation therapy, with doses adjusted to achieve targeted systematic exposure to eliminate individual differences due to variability in clearance. Patients with low-risk (LR) ALL were treated with an average dose of 2.5 g/m² and those with standard- or high-risk (SHR), 5.0 g/m² with dosages adjusted to achieve a plasma steady-state concentration of 33 μM or 65 μM, respectively. Courses of HDMTX were followed by standardized leucovorin rescue.

Figure 1:

Graphical representation of the timing of imaging and neurocognitive evaluations relative to the timeline of the therapy. Green circle represents start of therapy and red circle is end of therapy (EOT). Therapy is divided into three stages: Induction, Consolidation, and Continuation (120 weeks for females and 146 weeks for males). Both males and females were imaged and underwent neurocognitive testing at week 120 for consistency. Orange rounded boxes represent courses of high-dose methotrexate (HDMTX) that were administered as one course at 1.0 g/m² on day 1 of remission induction, and four additional courses given as 2.5 or 5.0 g/m² depending on risk stratification. Blue boxes represent the four MR imaging time points: upon completion of 6 weeks of remission induction (MR1), following consolidation treatment on week 7 (MR2), week 48 (MR3) and week 120 (MR4) of continuation treatment. Gold pentagons represent neurocognitive testing (Ψ) at EOT and 2-year follow-up.

Continuation treatment included weekly intravenous MTX at 40 mg/m² together with daily mercaptopurine for 3 weeks, followed by pulse therapy with vincristine plus dexamethasone at week 4. Continuation therapy spanned 120 weeks for girls and 146 weeks for boys and was interrupted by two reinduction treatments. Both males and females were imaged at week 120 for consistency. Therefore, the last patient was imaged in June 2010.

2.B. Quantitative MR Imaging and Diagnosis of LE

Longitudinal MR data were acquired at four time points: upon completion of 6 weeks of remission induction (MR1), week 7 (MR2) following consolidation treatment with 4 courses of HDMTX, week 48 (MR3) and week 120 (MR4) of continuation treatment (end-therapy time point). Patients demonstrated no neurologic symptoms at the time of imaging regardless of the presence or absence of LE.

All MR images were 4-mm-thick contiguous axial-imaging datasets collected on a 1.5-T Vision whole-body unit (Siemens Medical Systems, Iselin, NJ). T1-weighted images were acquired with a multi-echo inversion-recovery sequence (TR=8000ms, TE=29ms, TI=300ms, NEX=1, 7 echoes). Since LE is best detected by a T2-weighted sequence, preferably with cerebrospinal fluid (CSF) attenuation, a fluid attenuated inversion recovery (FLAIR) image was collected with a multi-echo sequence (TR=9000ms, TE=119ms, TI=2470ms, NEX=1, 7 echoes). A dual spin-echo sequence was used to acquire proton density and T2-weighted images simultaneously (TR=3500ms, TE1=17ms, TE2=102ms, NEX=1, 5 echoes). Diffusion-weighted imaging was not acquired on this protocol. Using an automated segmentation routine based on artificial neural networks, tissue maps were obtained for CSF, gray matter (GM), NAWM, and LE volumes (25). Variable inversion T1 (TR=5000ms, TE=600ms, NEX=1, TI=100/500/900/2324ms) and multiple echo T2 (TR=2000ms, NEX=1, 16 echoes sampled every 22.5ms) images were also acquired for two sections at the level of the basal ganglia and the centrum semiovale and fit with mono-exponential models to produce quantitative relaxation maps. This is very limited coverage and was designed to minimize acquisition time and still sample the periventricular and centrum semiovale white matter.

Datasets were read retrospectively by two neuroradiologists (FHL and NDS), each with over 20 years of experience. The neuroradiologists retrospectively assessed each patient’s longitudinal dataset for the presence of LE. LE was diagnosed according to radiographic criteria of Common Terminology Criteria for Adverse Events Version 4.0. None of the patients exhibited any neurologic symptoms at the time of imaging. Areas of LE ranged from small focal or subtle diffuse T2/FLAIR hyperintensities involving primarily periventricular white matter (Figure 2) to more extensive T2/FLAIR hyperintensities extending into centrum semiovale and could occupy more than one-third of the total white matter (Figure 3). Since the periventricular white matter is frequently involved and given the young age of patients, care was taken to differentiate therapy-induced white matter changes from normal terminal zones of myelination relying on the experience of the radiologists. Due to the large doses of corticosteroids used in therapy, increases in subarachnoid spaces or ventriculomegaly were not considered in the diagnosis of LE.

Figure 2:

T2-weighted images demonstrating appearance of more subtle leukoencephalopathy and its phenotypical expression over the course of therapy as either transient (top row) or chronic (lower row). The two images on the left are from early in therapy at the MR2 time point while the two images on the right are from end of therapy at the MR4 time point.

Figure 3:

A transverse T2-weighted fluid attenuated inversion recovery (FLAIR) image demonstrating avid leukoencephalopathy within the white matter of the centrum semiovale sparing the U-fibers

2.C. Neurocognitive Assessments

Assessments were conducted at week 120 of continuation therapy (end-therapy time point) and two years later at follow-up using age-standardized measures with demonstrated reliability and validity. Analyses focused on performance at two-year follow-up and change in performance from end of therapy to follow-up. Therefore, the last patient was assessed in October 2012. Patients ≥ 6 years of age at the time of the evaluation, completed measures of estimated IQ, working memory, processing speed (26, 27), and a computerized sustained attention measure (28), which yields scores for omissions, reaction time, variability, vigilance, and risk taking (as measured by β). Caregivers of patients ≥ 3 years of age at the time of the evaluation completed standardized ratings of attention and behavior in daily life (conduct, learning, and impulsivity) (29). Memory (as measured by the California Verbal Learning Test’s List A Total Recall) and academic skills (reading, spelling, math) were assessed in patients ≥ 6 years of age at the time of the evaluation (30–32). Patients had to meet these age requirements to be evaluated with these measures. All assessments were administered by master’s-level psychological examiners under the supervision of a licensed clinical psychologist. A previous report of these measures from this patient cohort identified that a significantly higher proportion of patients performed more than one standard deviation worse than normative expectations on omissions and risk-taking scores on a computerized measure of attention, total recall on the verbal list learning measure, processing speed on the intelligence measure, all three measures of academic performance, and all three indices on the caregiver report measure at the two-year follow-up (23). Based on these findings, we restricted our analysis to these specific measures.

2.D. Statistical Analyses

Chi-square tests were used to test the association between risk group and sex. Wilcoxon rank sum tests were used to compare age at diagnosis between risk groups and sex. Two-sided p-values were assessed. Generalized estimating equation (GEE) models with logit link were built to compare the prevalence of LE at different time points in all patients and in SHR and LR patients separately. Chi-square tests were used to test whether prevalence of LE between SHR and LR patients was different at each time point separately.

Neuroimaging measures of LE prevalence and the extent of white matter affected were modeled and compared between consecutive MR time points using GEE. Age at diagnosis was treated as a categorical variable with a threshold of 5 years of age (median age at diagnosis=5.35 years). This method fully utilized the longitudinal MRI data by considering the correlation between repeated measurements of individual patients. Individual GEE models for main effects of age at diagnosis, sex, and treatment risk arm were developed separately and interaction effects within the models were tested using chi-square tests based on the Wald statistics. A final comprehensive multivariable model was developed incorporating all main effects and their interactions. All models used backward model selection methods to eliminate insignificant factors (p>0.10) by the Wald test for type 3 analysis. However, if an interaction term was marginally significant (p<0.10), then the main effect variables were kept in the final model regardless of their individual significance.

CSF, GM, WM, and LE volumes were assessed longitudinally across the four time points by GEE modeling. Quantitative T1 and T2 relaxation rates of NAWM and LE from every MRI examination were plotted as a function of age at examination and assessed with a mono-exponential fit. This assessment did not take into consideration any regional variations within the brain.

General linear models were used to investigate the association between neurocognitive functions at two-year follow-up and neuroimaging metrics, including prevalence of LE, extent of LE at MR2 and extent of LE at MR4. Similar models were developed to investigate how those neuroimaging metrics are associated with the change in neurocognitive performance from end of therapy to the two-year follow-up. Due to the exploratory nature of this study and given the very selective choosing of cognitive and imaging variables of interest a priori based on the existing literature (resulting in a significant data reduction related to available cognitive and imaging variables), there was no need to control for multiple comparisons in the presented analyses. All statistical analyses were performed using SAS (version 9.3, SAS Institute, Inc.).

3. Results

3.A. Patient Characteristics

Due to clinical factors or MR image artifacts, we were unable to obtain all four time points from every patient. Table 1 details the number of examinations acquired at each time point and distribution by sex, age at diagnosis, and treatment risk arm. Since T-cell ALL, an SHR leukemia subtype, occurs more frequently in older males, and children ≤5 years of age are more likely to have LR leukemia, there is a substantial correlation between age at diagnosis, sex, and treatment risk arm across the whole cohort of 377 children (2). Sex was associated with both risk arm and age at diagnosis. There was a significantly higher proportion of females in the LR group (59.2%) and males in the SHR group (57.7%; p=0.001). The 208 males (mean=7.31±4.6 years) were significantly older than the 169 females (mean=6.46±4.7 years; p=0.014) at diagnosis. Across both risk groups, the 189 total SHR patients (mean=8.8±5.0 years) were significantly older than the 188 LR patients (mean=5.1±3.2 years; p<0.001) at diagnosis.

Table 1.

Number of MR examinations included in study at each time point. Number of exams are also categorized by age at diagnosis, sex, and risk arm on the protocol. Within each category, the ratio is also reported.

Data	MR1	MR2	MR3	MR4
All Patients	284	324	305	306
Age at Diagnosis
≤5 years	135	151	148	151
>5 years	149	173	157	155
Ratio >5 : ≤5 years	1.1	1.1	1.1	1.0
Sex
Female	127	150	132	140
Male	157	174	173	166
Ratio Male : Female	1.2	1.2	1.3	1.2
Risk Arm
Low Risk (LR)	150	169	164	166
Standard/High Risk (SHR)	134	155	141	140
Ratio LR : SHR	1.1	1.1	1.2	1.2

3.B. Prevalence of LE

Prevalence of LE was defined as the presence of LE at any of the four imaging time points. Similar to previous reports (7), LE prevalence significantly increased to its maximum at MR2 (74/324; 22.84%, p<0.001) compared to MR1 (22/284; 7.75%). Subsequently, LE prevalence significantly decreased during continuation therapy at MR3 (49/305; 16.07%, p=0.003) and MR4 (39/306; 12.7%, p<0.001) compared to MR2, regardless of risk arm, age at diagnosis or sex.

The prevalence of LE was higher in SHR patients relative to LR patients at MR2, MR3, and MR4 but was only significant at MR4 (8.4% vs.17.9%; p=0.016) as demonstrated in Figure 4A. For both SHR and LR patients, the prevalence of LE increased at MR2 compared with MR1 (SHR: 6.7% to 25.8%, LR: 8.7% to 20.1%, both: p<0.001). The prevalence then significantly decreased in the LR patients at MR3 LR: 20.1% to 13.4%, p<0.001) and in both groups at MR4 (SHR: 25.8% to 17.9%, p=0.035, LR: 20.1% to 8.4%, p<0.001) compared with MR2.

Figure 4:

Prevalence of leukoencephalopathy (LE) at each of the four imaging time points throughout therapy (MR1-MR4). Prevalence of LE shown as a function of (A) treatment risk arm (bars: black = standard/high risk [SHR] patients, hatched fill = low risk [LR] patients), (B) age at diagnosis (bars: black = >5 years old [Older], hatched fill = ≤5 years old [Younger]), and (C) sex (bars: black = male, hatched fill = female). P-values represent significant differences in LE prevalence in groups between time points.

Prevalence of LE was substantially higher in younger patients (≤ 5 years old at diagnosis) relative to older patients (> 5 years old at diagnosis) at MR2 but decreased at a faster rate such that by MR4 younger patients had a lower prevalence of LE than older patients as demonstrated in Figure 4B. In both age groups, LE prevalence significantly increased at MR2 compared with MR1 (older: 4.0% to 20.2%, younger: 11.9% to 25.8%, both: p<0.001). The prevalence then significantly decreased in younger patients at both MR3 (25.8% to 15.5%, p<0.001) and MR4 (25.8% to 9.9% p<0.001), compared with MR2. Prevalence in older patients did not significantly change from MR2 to end of therapy.

Males had a substantially but not significantly higher prevalence of LE relative to females at MR2, MR3, and MR4, which resolved at a slower rate than in females as shown in Figure 4C. For both sexes, LE prevalence signficantly increased at MR2 compared with MR1 (Males: 5.1% to 26.4%, p<0.001, Females: 11.0% to 18.7%, p=0.013), then significantly decreased at MR3 (Males: 26.4% to 18.5%, p=0.003, Females: 18.7% to 12.9%, p=0.032) and MR4 (Males: 26.4% to 15.1%, p<0.001, Females: 18.7% to 10.0%, p=0.003), compared with MR2.

Furthermore, a final comprehensive multivariable model (see Table, Supplemental Digital Content 1) was developed including time point, risk, sex, age at diagnosis, and their interactions. The Wald statistics for type 3 analysis of this model showed that time point (p<0.001), age at diagnosis (p=0.021), treatment risk arm (p=0.001), sex (p=0.087), interaction of age and time point (p=0.028) and interaction of treatment risk arm with both sex (p<0.001) and age (p=0.001) were influential factors associated with probability of developing LE.

3.C. Volumetry of Cortical Atrophy and Extent of White Matter Affected by LE

CSF, GM, WM, and LE volumes were assessed longitudinally across the four time points. Cerebral atrophy was evident in both the size of the ventricles and widening of the sulci early in therapy. When analyzed across all patients, quantitative CSF volumes relative to intracranial volume (ICV) significantly decreased between each of the four MR time points in this study ([CSF/ICV]; MR1=10.9%, MR2=8.7%, MR3=7.0%, MR4=5.9%, X²=884.39, p<0.001). These changes indicate decreasing cortical atrophy, which is likely associated with tapering of corticosteroids throughout the course of therapy.

Extent of LE was quantified by the volume of LE relative to the total WM volume, including both NAWM and LE, and was evaluated at all four time points in therapy. The extent of LE significantly increased at MR2 (p<0.001), compared to MR1, and subsequently, significantly decreased during continuation therapy at MR3 (p<0.001) and MR4 (p<0.001), compared to MR2.

Patients receiving more intensive therapy on the SHR risk arm demonstrated significantly more extensive LE at MR2 (p=0.040), MR3 (p<0.001) and MR4 (p=0.006) as can be observed in Figure 5A. For both SHR and LR patients, the extent of LE significantly increased at MR2 compared with MR1 (SHR: 2.9% to 9.5%, p<0.001, LR: 3.9% to 6.5%, p=0.004). Subsequently, the extent of LE then significantly decreased at MR3 (SHR: 9.5% to 7.6%, p=0.021, LR: 6.5% to 3.6%, p<0.001) and MR4 (SHR: 9.5% to 6.5%, p=0.008, LR: 6.5% to 3.6%, p<0.001) relative to MR2.

Figure 5:

Extent of white matter affected by leukoencephalopathy (LE) relative to total white matter at each of the four imaging time points (MR1-MR4) throughout therapy. Average extent shown as a function of (A) treatment risk arm (bars: black = standard/high risk [SHR] patients, hatched fill = low risk [LR] patients), (B) age at diagnosis (bars: black = >5 years old [Older], hatched fill = ≤5 years old [Younger]), (C) sex (bars: black = male, hatched fill = female), and (D) LE phenotype (bars: black = chronic LE patients, hatched fill = transient LE patients). P-values represent significant differences in the extent of LE in groups between time points.

Regardless of age, the extent of WM affected was similar at MR2. However, younger patients rapidly recovered over the next two years, while older patients maintained a more extensive involvement even at MR4 as demonstrated in Figure 5B. In both age groups, LE extent significantly increased at MR2 compared with MR1 (older: 2.3% to 8.0%, younger: 3.8% to 8.2%, both: p<0.001). The extent then significantly decreased in younger patients at both MR3 (8.2% to 3.5%, p<0.001) and MR4 (8.2% to 3.0%, p<0.001), compared with MR2. Extent of white matter affected in older patients did not significantly change from MR2 to end of therapy.

While the proportion of WM affected by LE in male and female patients were approximately equal at MR2, the extent of LE decreased more rapidly in female patients leaving male patients with more extensive changes at MR4 as seen in Figure 5C. In both groups, the extent of LE significantly increased at MR2 compared with MR1 (Male: 4.1% to 8.2%, p<0.001, Female: 3.0% to 7.9%, p=0.001). Thereafter, both groups significantly decreased at MR3 (Male: 8.2% to 6.7%, p=0.007, Female: 7.9% to 4.0%, p<0.001) and MR4 (Male: 8.2% to 6.1%, p=0.002, Female: 7.9% to 4.4%, p=0.020), compared with MR2.

Furthermore, a final comprehensive multivariable model (see Table, Supplemental Digital Content 2) was developed including time point, risk, sex, age at diagnosis, and their interactions. The Wald statistics for type 3 analysis of this model showed that time point (p=0.004), treatment risk arm (p=0.014), sex (p=0.084), and interaction of age and time point (p=0.092) were influential factors associated with the proportion of WM affected by LE.

The phenotypical expression of LE is characterized by its temporal evolution during therapy; it is either transient, defined as the presence of LE at any point in therapy but no LE on MR4, or chronic, defined as the presence of LE at MR4 regardless of when it first appeared. Only those patients that developed LE earlier in therapy and had imaging at MR4, 73 patients (39 chronic / 34 transient), could be evaluated. The extent of WM affected differed between chronic and transient phenotypes with more extensive WM involvement being associated with the chronic phenotype as appreciated in Figure 5D. For patients with the chronic LE phenotype, the extent of LE significantly increased at MR2 (3.7% to 9.7%, p<0.001), compared with MR1, and subsequently significantly decreased at MR3 (9.7% to 6.1%, p<0.001), compared with MR2. At end of therapy, MR4, patients with chronic LE still had 5.5% of their white matter affected.

3.D. Relaxometry of LE

Measurements of T1 and T2 relaxation rates from all patients across all four MR imaging time points included 1187 examinations, of which 175 had measurable regions of LE. Quantitative T1 and T2 rates in NAWM demonstrated significant reductions with age at examination (both: p<0.001). Quantitative relaxation rates were significantly longer in LE compared to NAWM in the same patient on the same examination (T1: p<0.001; T2: p<0.001) but were more easily separated from NAWM values on T2 than on T1 maps (Figure 6). T1 and T2 intensity (elevation in LE relaxation times relative to NAWM) was not significantly different by age at diagnosis, sex, risk group or LE phenotype.

Figure 6:

Relaxation times of normal appearing white matter (green dots) and leukoencephalopathy (gold dots) in all MRI examinations as a function of age at examination. Quantitative (A) T1 and (B) T2 relaxation times are shown.

3.E. Relating Neuroimaging During Therapy to Later Neurocognitive Performance

To investigate the relationships between neuroimaging metrics during therapy and neurocognitive performance after completion of therapy, we focused on early prevalence of LE (at either MR1 or MR2), extent of LE at MR2, and extent of LE at MR4 and evaluated the association of each of these imaging metrics with neurocognitive performance at the two-year follow-up. Selection of the early imaging metrics were driven by the desire to associate the prevalence and extent of LE early in therapy with much later neurocognitive performance in the hope of identifying patients that may benefit from target interventions. The inclusion of the MR4 time point was an alternative if the early imaging markers did not reach significance. Neither early prevalence nor extent of LE at MR2 was significantly associated with any of the performance or rater-based measures. Of the 174 patients that had imaging at MR4 and completed the academic performance measure at the two-year follow-up, only 19 (10.9%) had LE. However, even with this reduced sample size, greater extent of LE at MR4 was significantly associated with performance at the two-year follow-up on measures of reading (p=0.004), spelling (p=0.003) and mathematics (p=0.019) as demonstrated in Figure 7.

Figure 7:

Extent of white matter affected by leukoencephalopathy (LE) at the completion of therapy (imaging time point MR4) and its association with academic performance measures of (A) reading (p=0.004), (B) spelling (p=0.003), and (C) mathematics (p=0.019) on WIAT at the two-year follow-up.

Since decreasing neurocognitive performance from the end of therapy to a two-year follow-up in these patients has previously been demonstrated (23), we next assessed the association of early prevalence of LE defined as LE at either MR1 or MR2, extent of LE at MR2, and extent of LE at MR4 with these differences. Parent reports at end of therapy and at two-year follow-up were available for 150 patients (114 with NAWM and 36 with early prevalence of LE). Early prevalence of LE was found to be significantly associated with increasing parent reports of conduct problems (NAWM: mean change = −2.8, early prevalence of LE: mean change = 1.1, p=0.039) and learning difficulties (NAWM: mean change = −1.3, early prevalence of LE: mean change = 4.9, p=0.036). Of the 81 patients that had imaging at MR2 and completed the processing speed measure at end of therapy and at two-year follow-up, only 18 (22.2%) had LE. A larger extent of LE at MR2 was significantly associated with decreasing processing speed (p=0.003) after therapy as shown in Figure 8A. Similarly, of the patients that had imaging at MR4 and completed the attention measures (N=106) and memory measure (N=107) at end of therapy and at two-year follow-up, only 14 (13.2%) had LE. A larger extent of LE at MR4 was significantly associated with increasing problems with attention (Omissions (p=0.045); Beta (p=0.015)) and memory (List A Total Recall (p=0.010)) as shown in Figures 8B-8D.

Figure 8:

Extent of white matter affected by leukoencephalopathy (LE) and its association with differences in neurocognitive performance from the end of therapy to the two-year follow-up evaluation. (A) A greater extent of LE early in therapy, following consolidation treatment (imaging time point MR2), was significantly associated with decreasing processing speed (p=0.003) on WISC/WAIS. A greater extent of LE at the completion of therapy (imaging time point MR4) was significantly associated with increasing attention problems given by (B) Omissions (p=0.045) and (C) Beta (p=0.015) scores on CPT, and poorer memory given by (D) List A Total Recall (p=0.010) scores on CVLT.

4. Discussion

This study systematically examined the prevalence, extent, and intensity of LE during therapy and the association of these early imaging markers with later neurocognitive performance in a large single-institution cohort of risk-adapted, uniformly treated children with ALL. LE was most prevalent (22.8%) at MR2, shortly after the completion of upfront intrathecal therapy for CNS prophylaxis and consolidation therapy with four courses of HDMTX. LE occurred more frequently and involved a larger extent of the white matter in males than in females and among patients treated with the more intensive SHR than the LR regimen. These findings likely reflect the known higher prevalence of the more aggressive T-cell leukemia subtype in males and the more intensive CNS-directed therapy SHR patients receive (2).

Patients treated on the less intensive LR arm were significantly younger than those on the more intensive SHR arm, in agreement with previously published literature (2). Prevalence of LE was substantially higher in younger patients relative to older patients at MR2 but decreased at a faster rate such that by MR4 younger patients had a lower prevalence of LE than older patients. Likewise, the proportion of white matter affected at MR2 was similar regardless of age but resolved in younger patients over the next two years while older patients maintained a more extensive involvement even at MR4. Taken together, these results demonstrated that younger patients treated with less intense therapy were more susceptible to developing LE early in treatment, but most returned to having no LE by the end of therapy, which is consistent with a greater degree of brain plasticity that facilitates repair of less intense damage (33). Conversely, older patients treated on the more intense SHR arm presumably have more severe WM damage and less ability to repair it, resulting in a higher rate of chronic LE.

For the first time, the development and extent of LE during therapy were associated with clinical outcomes in the neurocognitive performance of survivors two years after the completion of therapy. A greater extent of LE at the end of therapy (MR4) was significantly associated with decreased performance on measures of reading, spelling, and mathematics two years later. While decreasing neurocognitive performance from the end of therapy to a two-year follow-up in these patients has previously been demonstrated (23), this study showed multiple novel associations between this decline and the presentation of LE at specific time points during therapy. Early prevalence of LE, LE at either MR1 or MR2, was significantly associated with increasing parent reports of conduct problems and learning difficulties. Further, a larger extent of LE during early therapy (at MR2) was significantly associated with decreasing processing speed, while a larger extent of LE at the end of therapy (at MR4) was significantly associated with increasing deficits in attention and memory. These novel associations between LE and cognition assist in increasing understanding of the etiology of cognitive late effects and may promote refinement in treatment and/or development of interventions that could reduce risk. While the spatial localization of the LE is predominantly in the centrum semiovale and periventricular white matter, (14) the diffuse nature of the lesions makes association of individual lesions with specific neurocognitive performance infeasible.

These associations between the prevalence and extent of LE during therapy with later neurocognitive performance are consistent with the localization of the LE within the deep white matter of the frontal, corona radiata and periventricular regions (14). In long-term survivors of childhood ALL, LE has been associated with damage to the white matter microstructure and neurocognitive impairment (34, 35). A subset of 46 patients that had LE during therapy on the current study were imaged again an average of 5 years after therapy and found that 78% (36/46) continued to have LE (19). Damage in these specific regions is likely to have a significant impact on frontal mediated cognitive functions in attention and working memory and lead to slower processing speed.

T1 and T2 relaxation times in LE were longer than in NAWM. However, T2 relaxation rates better differentiated NAWM from LE. Longer T1 and T2 relaxation times may reflect biological underpinnings of LE related to MTX. In long-term survivors of childhood ALL, higher plasma MTX concentrations have been associated with damage to the white matter microstructure in the frontostriatal tracts and with increased rates of below-normal neurocognitive performance (36). Methotrexate-based deficiencies in iron levels and the folate-vitamin B12-methylation pathway cause hyper-homocysteinemia and result in deficient myelin synthesis (3–6, 37–39). LE is identified by regional hyper-intensities on T2W images, which had longer T2 relaxation times, possibly resulting from iron deficiencies caused by decreased levels of folate. Iron is essential to myelin synthesis (40–42). A decrease in iron and deficient metabolism involving the methyl transfer pathway are known to cause a decrease in the production of myelin proteins and lipids, resulting in hypo-myelination and inadequate myelin compaction (43, 44). Longer T1 relaxation times in LE may be linked to a decrease in magnetic interactions of hydrogen nuclei resulting from decreased hydrophobic lipids in LE. Increased interstitial water content, reflecting less structured WM due to decreased myelin compaction, may also be associated with longer T1 relaxation times in LE. Longer T1 and T2 relaxation times in LE may, therefore, reflect causation, at least in part, by MTX-induced reduction in iron levels and disruption of the folate-vitamin B12-methylation pathway. Future work should investigate these biological underpinnings using diffusion tensor imaging, which is more sensitive and specific to WM microstructure.

Diffusion-weighted imaging (DWI) has been used to image ALL patients presenting with sudden onset of a central neurological syndrome within days of intrathecal MTX (45, 46). This syndrome can include slurred speech, emotional lability, and hemiparesis and usually occurs within a few days of intrathecal MTX administration. DWI in these cases display a restricted diffusion pattern reflecting cytotoxic edema within cerebral white matter. This finding is most consistent with a reversible metabolic derangement, rather than ischemia. In contrast, none of the patients in this study exhibited any neurologic symptoms at the time of imaging. Based on studies in ALL survivors, DWI in these patients would most likely exhibit elevated diffusion (34). However, this hypothesis would need to be tested prospectively.

There are limitations of this study. This study reports on patients treated from 2000–2010 and therapy has continued to evolve including the most recent changes, which now included risk-stratification and treatment based on genetic phenotypes (47, 48). However, high-dose intravenous methotrexate and triple IT chemotherapy with methotrexate, cytarabine, and hydrocortisone remain the mainstay of CNS-directed prophylactic therapy in children treated for ALL. Time points in this study were chosen based on a previous study that included more frequent imaging and determined that the maximum prevalence occurred immediately following Consolidation (7). Patients have a “break” of 6 weeks following Consolidation to recover from this intense therapy before continuing to receive the first of two re-inductions during Continuation (see Figure 1). There may exist a more optimal time to image but determining this timing must be made in consideration of the patients, their families, and the limited resources available. The present analysis did not control for multiple comparisons due to the exploratory nature of the study and given the very selective choosing of cognitive and imaging variables of interest a priori. It has been demonstrated previously that a strict adjustment across an entire body of work is less critical for exploratory studies as subsequent prospective studies with preplanned hypotheses may be conducted in the future to confirm the observed associations (49–51). Furthermore, bar charts between groups and time points were presented for each analysis so the reader can see how different the groups are and can use their own judgment about the relative weight of the conclusions rather than relying on an arbitrarily thresholded p-value. The wide age range of patients limits the specificity of results for ALL patients of all ages; however, the enrollment of young patients was restricted to those who were one year of age and older to decrease the difficulty in distinguishing LE from neurodevelopmentally-driven, partially myelinated or unmyelinated WM. The study’s ability to investigate longitudinal changes across all four MR time points was affected by missing data for some patients. Elongated T1 and T2 relaxation times in LE may reflect the effect of MTX, but the contribution of other agents used during the course of therapy should also be considered. Results related to LE phenotype outcomes were limited to 73 patients who developed LE during therapy and had an evaluable scan at the final imaging time point. Finally, the effect of LE on WM development is not fully understood. It is possible that remyelination of existing tracts or the development of aberrant tracts occurs as a compensatory mechanism in response to regional LE.

Given the results of the current study, a more contemporary imaging protocol can be designed to take advantage of more recent technological developments. Rather than collecting relatively lengthy independent sequences for T1, T2, PD, and FLAIR imaging, a QRAPMASTER (quantification of relaxation times and proton density by multi-echo acquisition of a saturation recovery using turbo spin-echo readout) pulse sequence, which is a multi-slice, multi-echo, and multi-saturation delay acquisition sequence could be acquired requiring only 5 minutes (52). Synthetic MRI can reconstruct the corresponding T1, T2, PD, and FLAIR imaging contrast as well as whole head T1 and T2 relaxation maps and myelin water maps all from this one acquisition. Unfortunately, there are some limitations such as inferior image quality (lower contrast-to-noise ratio) in the synthetic FLAIR images as well as fluid pulsation artifacts in synthetic T2 weighted images (53). These may necessitate the acquisition of a separate FLAIR weighted image that could be acquired with a quite gradient acquisition requiring only 3 minutes and 38 seconds to be more compatible with imaging of young subjects (54). Recent advances in imaging of other white matter lesions has demonstrated the utility of double inversion recovery sequences, which attenuate the normal white matter signal while maintaining the T2 weighting of the lesions. Use of compressed sensing means this sequence can be acquired in just over 3 minutes (55). Lastly, the segmentation of LE could be automated using a computational pipeline developed for the quantification of multiple sclerosis (56). Unfortunately, while this approach is highly correlated with manual segmentations, it still has some limitations such as a limited lesion detection rate of only 80% and a median false positive rate of 33% (56).

In summary, this study showed that LE was most prevalent at the completion of consolidation therapy. Patients that were older at diagnosis received more intense therapy, and males were at a greater risk for developing LE that was more likely to involve a larger extent of the white matter. More extensive white matter involvement across all patients was associated with the chronic LE phenotype and, when observed at the end of therapy, was significantly associated with decreased academic performance years after treatment. Furthermore, decreasing neurocognitive performance from the end of therapy to a two-year follow-up showed that a larger extent of LE early in therapy was significantly associated with decreasing processing speed, while a greater extent of LE at the end of therapy was significantly associated with increasing problems with attention and memory. T2 relaxation rates best differentiated NAWM from LE and were best appreciated using fluid-attenuated T2-wieghted sequences. Overall, asymptomatic LE during therapy can be seen in almost a quarter of patients during therapy, involve as much as 10% of WM volume, and is associated with decreasing neurocognitive performance in survivors.