Impairment in circadian activity rhythms occurs during dexamethasone therapy in children with leukemia

Abstract

Background:

Circadian rhythm disturbances in adults with cancer are associated with fatigue, time to relapse, and death. This study of circadian activity rhythms (CAR) of children with acute lymphocytic leukemia (ALL) on continuation chemotherapy aimed to describe CAR before and after starting dexamethasone, and to determine whether fatigue was associated with less robust CAR.

Procedure:

This was a secondary analysis of data from a multi-institutional study in which children with ALL aged 5–18 years wore an actigraph for 10 consecutive 24-hour periods, 5 before and 5 during dexamethasone therapy. CAR parameters measured by actigraphy were calculated for each 5-day period, including peak activity, MESOR, amplitude, acrophase and circadian quotient. Fatigue was measured on study days 2, 5, 7 and 10 by parent-report and self-report for children ≥7 years.

Results:

Eighty-two children qualified for CAR analysis, and 87 for analysis of daily peak activity patterns and fatigue. Mean age was 8.8±3.3 years. Peak activity, MESOR and amplitude significantly decreased during dexamethasone therapy. Children on high-dose dexamethasone (8 or 12 mg/m²/d) had significantly higher (better, or more robust) values of several CAR parameters than those on low-dose (6 mg/m²/d). There was a significant trend of decreasing daily pattern of peak activity during dexamethasone therapy only. Fatigue increased across the study and was associated with decreasing CAR peak activity, MESOR and amplitude.

Conclusions:

Dexamethasone initiation was associated with a decrease in several CAR parameters, and a significant decrease in the trend of daily peak activity. Fatigue was associated with less robust CAR.

INTRODUCTION

The robustness of our circadian rhythms indicates their ability to maintain strength and stability in the face of physiologic and environmental challenges. Adult cancer patients often experience dampening, or decreased robustness, of their circadian rhythms [1–3]. These patients experience greater fatigue, poorer quality of life (QoL), and decreased time to relapse and death than patients with robust rhythms [4–6]. Multiple factors contribute to disturbance of circadian rhythms in these patients including the cancer, its treatment, and associated pain and changes in psychological state [7–10]. Additionally, cancer treatment can cause lifestyle changes that decrease exposure to powerful environmental time cues, called zeitgebers, which help maintain our circadian rhythms. Zeitgebers include exposure to social interactions, eating, exercise and most importantly daylight [11–13].

The rhythm of a person’s locomotor activity across 24-hour periods is one type of circadian rhythm. These circadian activity rhythms (CAR) can be measured using a wrist-worn device called an actigraph, which measures movement over time and is used to estimate sleep-wake activity and CAR. To date, CAR research in cancer has focused on adults. No studies have evaluated CAR in either healthy children or in children with cancer. Yet given the negative consequences of circadian disruption and the potential for improving outcomes by supporting CAR, circadian rhythms in children demand investigation. The primary aim of this study was to describe CAR of children with acute lymphoblastic leukemia (ALL) before and during a pulse of dexamethasone. A secondary aim was to determine whether less robust CAR was associated with fatigue, a common experience of children with cancer. We hypothesized that initiation of dexamethasone would be associated with decreased robustness of CAR, and that less robust CAR parameters would be associated with greater fatigue.

METHODS

Study Design

This was a secondary analysis of data from a previously described original, prospective study of sleep and fatigue in children and adolescents with ALL [14–17]. Children included in the present study were a subsample of participants from that ‘parent’ study.

Participants

Children taking part in the parent study were recruited from three pediatric oncology centers: St. Jude Children’s Research Hospital (St. Jude; Memphis, TN, USA), Texas Children’s Cancer Center (Houston, TX, USA) and the Hospital for Sick Children (Toronto, Canada). Children at St. Jude were treated on the Total XV protocol. Children at the other institutions were treated on Children’s Oncology Group (COG) protocols 9904 or 9905. By protocol, children were excluded if they had CNS disease at diagnosis or relapsed during treatment, and all low-risk children were less than 10 years old. No children received radiation. Inclusion criteria for the parent study were ages 5–18 years, diagnosed with low-risk or standard-risk ALL, and undergoing continuation chemotherapy during weeks 50–76 of their protocol. Children were evaluated during continuation therapy because it is a period of relative medical stability, they could participate while living in their own homes where the rhythm of daily activity and sleep would be relatively normal, and because dexamethasone, although instrumental in improving survival in children with ALL, is associated with adverse effects that disrupt the lives of children and their families during administration [18–21]. Dosing of dexamethasone on COG protocols was 6 mg/m²/day and on the Total XV protocol was 8 mg/m²/day for low-risk and 12 mg/m²/day for standard-risk ALL. An additional criterion for inclusion in this study was availability of actigraphy data before and during dexamethasone therapy.

The parent study was approved by the institutional review boards (IRBs) at each study site. This study was reviewed by the IRB at the University of Maryland, Baltimore and assigned exempt status.

Measures

Actigraphy

The study period was timed to measure actigraphy over 10 consecutive 24-hour periods, five prior to the start of a pulse of dexamethasone (pre-DEX, days 1–5) and five while taking dexamethasone (on-DEX, days 6–10). Actigraphy data were recorded with a Mini Motionlogger AAM-32 actigraph (Ambulatory Monitoring, Inc., Ardsley, NY) worn on the dominant wrist. Placement on the dominant wrist was recommended by our consultant (Dr. Avi Sadeh, Tel Aviv University, Tel Aviv, Israel) to best capture all activity levels. A daily sleep diary was maintained by parents concurrent with actigraphy measurement.

Circadian Activity Rhythms

From actigraphy output, parameters for CAR analyses were estimated by actigraphy software using cosinor analysis, which fits a cosine curve with a period of 24 hours to the raw activity count data [22]. A minimum of 72 hours of actigraphy data during each period, pre-DEX and on-DEX, were required for CAR analyses. The following five parameters were calculated. Peak activity is the highest activity count in a 24 hour period. Midline estimating statistic of rhythm (MESOR) is the halfway point between the peak and trough activity counts over each 24-hour period. Amplitude is the difference between the peak and trough activity. It represents the change in activity across 24 hours. Circadian quotient represents the strength of the CAR, calculated by dividing the amplitude by the MESOR. This normalizes the value of the amplitude and allows comparison between individuals [22]. Acrophase is the clock time at which peak activity occurs. Shifts toward a later acrophase suggest a delayed sleep phase, while shifts toward an earlier acrophase suggest an advanced sleep phase. Either phase shift can misalign the circadian clock from the 24-hour solar cycle, and potentially diminish an organism’s inherent ability to adapt behavior and physiological function to biologically advantageous times during the day [23]. Higher values of the first four parameters indicate more robust CAR. Figure 1 demonstrates the concept of robustness of CAR by comparing actograms, or pictorial representations of actigraphic activity data, of two children with sickle cell disease, one of whom had a robust CAR and another whose CAR was much less robust. Raw actigraphy output was not available to demonstrate this concept in our sample.

Figure 1.

Examples of circadian activity rhythms (CAR) across 5 days, measured by actigraphy, in two children with a non-cancer chronic illness (sickle cell disease). (A) Depicts a child with robust CAR, noted by high activity counts (black lines) during the day and low activity counts at night (high within-day variability), and robust consistency in the rest-activity pattern across days (high between-day stability). (B) Depicts a child with less robust CAR, noted by lower daytime activity, higher nighttime activity, indistinct sleep periods and a poorly reproducible rest-activity pattern across days. Activity counts are set to the same activity scale (3000) for both children.

Fatigue

On study days 2, 5, 7 and 10, fatigue was self-reported by children 7 years and older, and parent-reported for all children and adolescents, using validated questionnaires. These measures were completed at home and mailed to investigators in prepaid mailers. Data related to fatigue from the parent study have previously been reported [14]. Fatigue measures included the following scales. On all scales, higher scores indicated greater fatigue. The Child Fatigue Scale is a 14-item self-report instrument for children aged 7–12 years that scores the presence and intensity of cancer-related fatigue symptoms on a 5-point Likert scale. Scores range from 0–70. Cronbach’s alpha in the validation study ranged from 0.72–0.81 [24]. The Adolescent Fatigue Scale is a 13-item revised self-report instrument for adolescents aged 13–18 years that scores the intensity of cancer-related fatigue on a 5-point Likert scale. Scores range from 13–65. Cronbach’s alpha in the validation study ranged from 0.89–0.95 [25]. The Parent Fatigue Scale is a 17-item parent-report of their child’s cancer-related fatigue on a 5-point Likert scale. Scores range from 17–85. Cronbach’s alpha in the validation study ranged from 0.91–0.92 [24].

Statistical Analysis

Descriptive statistics included means and standard deviations calculated for continuous variables and proportions for categorical variables. Within-subject differences in CAR parameters across time (pre-DEX compared to on-DEX periods) were tested using paired t-tests. For group comparisons, age was dichotomized into children (ages 5–12 years) and adolescents (ages 13–18 years) based on the fatigue assessment instrument they completed. Dexamethasone dose was dichotomized into low-dose (6 mg/m²) and high-dose (8 or 12 mg/m²) (DEX groups). Between-group differences were tested with independent t-tests. In separate repeated-measures ANOVA models for each CAR parameter, interactions were tested between time and age, and time and DEX group.

Piecewise linear mixed effects models were used to test trends in daily activity patterns across pre-DEX and on-DEX periods, differences in trends between periods, and interactions between trends and age, and trends and DEX group. By comparing trends or slopes, an estimate of treatment effect can be obtained [26]. Peak activity was evaluated as the surrogate parameter for these analyses. Time was centered at day 5, the day prior to the start of dexamethasone, and continuous predictors were centered at their grand means. Change in fatigue across the study, and its relationship to CAR, were tested with random effects linear mixed models. This technique captures inter-individual differences and intra-individual correlations between repeated measurements of each child, and allows for some missing data [27]. Analyses were carried out with SPSS 19.0® (IBM SPSS Statistics for Windows, Armonk, NY: IBM Corp), and SAS 9.3® (SAS Institute, Inc., 2012, Cary, NC). Significant level was set at p<0.05.

RESULTS

Characteristics of the sample are described in Table I. Of the 100 children in the parent study, 13 lacked actigraphy data due to equipment failure or noncompliance with use. Five additional children did not meet minimum actigraphy recording time of 72 hours during each 5-day period to perform CAR analysis. The final sample included 82 children for CAR analysis and 87 for analysis of daily activity patterns and fatigue.

Table I.

Demographic and clinical characteristics of the sample

Parameters	N (%)	Mean ± SD
Age (years)		8.8 ± 3.3
Age group:
5–12 years	76 (87.4)
13–17 years	11 (12.6)
Sex (male)	57 (65.5)
Race:
Caucasian	71 (81.6)
African American	11 (12.6)
Other	5 (5.7)
ALL risk category:
COG low-risk	12 (13.8)
COG standard-risk	25 (28.7)
St. Jude low-risk	23 (26.4)
St. Jude standard-risk	27 (31.0)
Child Fatigue score
Pre-DEX		23.0 ± 7.2
On-DEX		28.7 ± 9.2
Adolescent Fatigue score
Pre-DEX		23.0 ± 7.8
On-DEX		29.0 ± 11.3
Parent Fatigue score
Pre-DEX		34.1 ± 10.9
On-DEX		44.4 ± 13.1
Daily dexamethasone dose (mg/m2)		9.1 ± 4.4

ALL refers to acute lymphoblastic leukemia; COG, Children’s Oncology Group; On-DEX, study days during dexamethasone therapy (days 6–10); Pre-DEX, study days prior to the start of dexamethasone (days 1–5); (n=87).

Circadian Activity Rhythms

Differences in CAR parameters across time for the sample are shown in Table II. Peak activity, MESOR and amplitude were significantly lower during the on-DEX compared to the pre-DEX period (all p<0.001), and acrophase occurred significantly earlier on-DEX (p=0.043), phase advancing by 17 minutes compared to pre-DEX. There was no significant difference in the circadian quotient from pre-DEX to on-DEX.

Table II.

Mean circadian activity rhythm values Pre-DEX and On-DEX

Circadian parameters	Pre-DEX Mean(SD)	On-DEX Mean(SD)	p-value
Peak activity	248 (51)	220 (52)	<0.001
MESOR	134 (29)	119 (31)	<0.001
Amplitude	114 (24)	101 (26)	<0.001
Circadian Quotient	0.87 (0.12)	0.88 (0.15)	0.262
Acrophase*	14:52 (1:31)	14:35 (1:45)	0.043

^*Acrophase reports a clock time (hr:min, in 24-h format). On-DEX refers to study days during dexamethasone therapy; Pre-DEX, study days prior to the start of dexamethasone; (n = 82).

Group differences in CAR parameters showed that, compared to participants on low-dose, those on high-dose dexamethasone demonstrated significant changes in CAR for several parameters. These included higher on-DEX peak activity (p=0.021), pre-DEX and on-DEX MESOR (p=0.005 and p=0.025) and on-DEX amplitude (p=0.026); later pre-DEX acrophase (p=0.003); and lower pre-DEX circadian quotient (p=0.004). No significant differences were found on any CAR parameter between age groups or sexes. There were no significant interactions between time and age or time and DEX group.

Trends in Daily Activity Patterns across Study Days

Parameters in the unadjusted piecewise linear mixed effects model of trends in peak activity were pre-DEX days and on-DEX days. In separate models, interactions between on-DEX trend and age, and on-DEX trend and DEX group, were included to test whether age or DEX dose had modifying effects on the trend in peak activity. No interaction terms were significant, so none were included in the final model.

There was no significant change in trend of peak activity across pre-DEX days, either before or after adjustment for age and DEX group (Table III), suggesting that peak activity was relatively stable across the pre-DEX period. There was a significant trend for decreasing peak activity once dexamethasone was started (b =−10.4, p<0.001) and a significant difference in trends between pre-DEX and on-DEX periods (b_on-DEX - b_pre-DEX =−11.7, p<0.001 unadjusted model; b_on-DEX - b_pre-DEX =−11.8, p<0.001 adjusted model) using bootstrap simulation with 1,000 repetitions (power to detect differences between trends=1.0). The daily pattern of peak activity was similar between DEX groups, with a trend of decreasing peak activity beginning immediately on starting dexamethasone. However, children on high-dose maintained a higher mean peak activity across the study than children on low-dose dexamethasone (Figure 2).

Table III.

Piecewise mixed effect models of peak activity

	Unadjusted model		Adjusted model
Parameter	Estimate (SE)	p-value	Estimate (SE)	p-value
Intercept	250.9 (5.55)	<0.001	237.4 (16.0)	<0.001
Pre-DEX	1.33 (1.24)	0.289	1.34 (1.24)	0.283
On-DEX	−10.4 (1.09)	<0.001	−10.4 (1.09)	<0.001
DEX group	--	--	28.1 (10.25)	0.007
Age	--	--	−2.97 (1.57)	0.061

Dexamethasone (DEX) group: 6 mg/m² (reference group) versus 8 or 12 mg/m². Trend tests the difference in slope between pre-DEX days and on-DEX days; (n=87).

Figure 2.

Plot of the slopes of peak activity across 10 consecutive 24-hour periods. Peak activity remained relatively stable across pre-DEX days (study days 1–5), then decreased significantly across on-DEX days (study days 6–10). Each line represents the slope of one child’s peak activity (n=87). The solid, heavy black line represents the grand mean of peak activity, the dashed line represents children on low dose dexamethasone (6 mg/m²/day) and the dotted line represents children on high dose dexamethasone (8 or 12 mg/m²/day).

Circadian Activity Rhythms and Fatigue

Separate linear mixed models were used to test the association between each CAR parameter and fatigue scores on days 2, 5, 7 and 10, controlling for DEX group (Table IV). Parent-reported child fatigue across the study was negatively associated with peak activity (b=−0.067, p<0.001), MESOR (b=−0.13, p<0.001) and amplitude (b=−0.09, p<0.001) but not with acrophase or circadian quotient. Findings were similar for self-reported child and adolescent fatigue. Results suggested that as CAR decreased following the start of dexamethasone, fatigue became more acute.

Table IV.

Association between fatigue scores and circadian activity rhythm parameters

	Adolescent Fatigue Score		Child Fatigue Score		Parent Fatigue Score
	Estimate (SE)	p-value	Estimate (SE)	p-value	Estimate (SE)	p-value
Peak	−0.10 (0.03)	<0.001	−0.03 (0.01)	0.005	−0.067 (0.014)	<0.001
MESOR	−0.17 (0.05)	0.001	−0.06 (0.02)	0.002	−0.13 (0.024)	<0.001
Amplitude	−0.13 (0.04)	0.004	−0.04 (0.02)	0.067	−0.09 (0.025)	<0.001
Acrophase	0.00003 (0.0003)	0.9	0.0002 (0.0001)	0.189	0.0002 (0.00015)	0.117
Circadian quotient	−10.4 (7.64)	0.182	3.1 (2.9)	0.301	6.9 (3.9)	0.076

Random intercept linear mixed modeling was used to estimate the association between fatigue and each of the circadian activity rhythm parameters across study days 2, 5, 7 and 10, with adjustment for dexamethasone dose (6 mg/m² versus 8 or 12 mg/m²).

DISCUSSION

This study undertook to describe CAR of children aged 5–18 years with ALL who were on chemotherapy during a period preceding and then during a pulse of dexamethasone. When children were on dexamethasone, we found dampening of several CAR parameters including peak activity, MESOR and amplitude, and phase advance of acrophase compared to the pre-DEX period. Similar changes were seen in both age groups and sexes. Participants on both high-dose and low-dose dexamethasone experienced similar dampening of CAR during the period on-DEX compared to pre-DEX. Yet children on high-dose dexamethasone demonstrated more robust CAR across the study than children on low-dose. Finally, we found that decreasing CAR parameters peak activity, MESOR and amplitude were significantly associated with increasing fatigue across the study, controlling for DEX group.

The finding that chemotherapy in patients with cancer impairs CAR is not novel. In adult women with breast cancer receiving anthracycline-based chemotherapy, Savard et al. found that compared to baseline, all circadian parameters except acrophase were significantly impaired during week one of the first and fourth cycles of their chemotherapy [3]. While some recovery occurred during the second and third weeks, impairment was significantly greater during the fourth compared to the first cycle, indicating that repeated cycles of chemotherapy caused progressive impairment of CAR [3].

We found dampening of several circadian parameters following the start of dexamethasone therapy. In our previous analysis, dexamethasone pharmacokinetics was studied across 8 hours following administration of the first dose of a dexamethasone pulse. Vallence and colleagues found that as the area under the concentration curve of dexamethasone increased, sleep efficiency and actual sleep time significantly decreased, and nocturnal awakenings and wake after sleep onset significantly increased [16]. The effects of glucocorticoids on increased latency to sleep, decreased sleep efficiency and shortened total nocturnal sleep time have also been demonstrated in adults [20,28,29]. It follows that disturbed sleep associated with dexamethasone might make children sleepier during the day. Thus the peak, MESOR and amplitude of activity would likely decrease, consistent with our findings.

In light of this, our finding that children on high-dose dexamethasone had higher CAR parameters than children on low-dose is intriguing. It could be that, while sleep disturbance caused by dexamethasone has an overall effect of dampening CAR, it still exerts enough of a stimulating effect [21,30] that children on high-dose dexamethasone are able to maintain a higher activity level than those on low-dose. This finding is preliminary and requires replication to establish the true relationship between DEX and dysregulated CAR. Sleep loss caused by dexamethasone could lead to a paradoxical hyperactivity and behavioral dysregulation, a common finding in healthy children who are chronically sleep restricted [31,32]. Indeed, we began this study at the behest of parents who repeatedly described their child while on dexamethasone as ‘not my child,’ ‘moody’ and ‘disruptive of the family,’ and who described the impending on-DEX period in terms such as ‘we tell each other “get ready—the dex week is coming”.’

Fatigue is a common experience in adults and children with cancer. In adults, it presents before initiation of chemotherapy or radiation therapy, and persists throughout treatment and beyond [1,14,33,34]. Studies of adults with cancer have demonstrated an association between fatigue and disturbed circadian rhythms [35]. Roscoe et al., in a sample of women receiving their second and fourth cycle of chemotherapy for breast cancer, found that an increase in fatigue from the second to the fourth cycle was significantly associated with a decrease in autocorrelation, a measure of consistency of CAR across time [36]. In another study, Liu et al. found more severe cancer-related fatigue and disrupted CAR among breast cancer patients prior to chemotherapy than among controls, and more severe fatigue and disruption of CAR after four cycles of chemotherapy compared to baseline measures [37]. In our study, we demonstrated an inverse relationship between CAR parameters and fatigue, with decreasing robustness of CAR associated with significantly increasing fatigue.

The relationship between fatigue and dexamethasone is complex. Dexamethasone demonstrates an alerting effect in healthy adults [21,30], so it seems counterintuitive for children to experience fatigue while on this therapy. Yet it also disrupts sleep, as we and others have found [14,38]. Sleepiness, or the propensity to fall asleep due to sleep loss, is not the same as fatigue. Cancer-related fatigue is a feeling of tiredness to the point of exhaustion unrelated to activity, or out of proportion to the level of activity, that does not resolve with rest [39]. Fatigue can be physical (e.g. too tired to play) or cognitive (e.g. difficulty with attention or memory). It is a common experience in cancer, with over 90% of adult cancer patients reporting fatigue [40], yet its physiology is not understood. We can only speculate in our study that beyond the effect of the cancer itself, dexamethasone may alter physiological or psychological processes that contribute to the experience of fatigue. Mood disturbance, for example, often occurs on dexamethasone and also frequently co-occurs with fatigue [38]. Studies of fatigue in children with cancer and its causes are urgently needed. They should include repeated measurements of a broad scope of physiological and psychological variables that might influence the relationship between fatigue and dexamethasone such as emotional regulation, depression, neurocognition, hyperactivity, physical stamina and inflammatory markers, which may help to explain this association.

This study had several limitations. No normative values for CAR parameters exist for children or adolescents, thus we were unable to interpret our findings in light of CAR values typical of healthy children. Our sample size, although quite large for a study of pediatric cancer patients, was too small to perform subgroup analyses that might have allowed better identification of specific groups or personal characteristics that might predispose to the effects of dexamethasone. Particularly, we had few adolescents, and thus analyses including this age group must be considered exploratory. Children were on multiple medications at the time of the study, so we cannot fully exclude that medications other than dexamethasone influenced our findings. For example, vincristine was administered on day six, the day on which dexamethasone was started. However, vincristine dose did not differ between DEX dosing groups, and none of the five CAR parameters differed between children who did and did not receive vincristine, controlling for age, DEX dosing group and ALL risk group. So vincristine was unlikely to have influenced our results.

Based on our findings and the state of the science in pediatric cancer circadian rhythm research, we propose the following research priorities. First, we must define CAR in healthy children, to provide an appropriate comparison group for children undergoing cancer treatment. Second, changes in CAR after identification of cancer but prior to initiation of treatment should be studied to differentiate the effect of the cancer from the effect of its treatment. Although many pediatric cancer diagnoses, including ALL, are emergencies requiring immediate intervention, treatment of some cancers is delayed in order to better define the diagnosis and appropriate treatment. This delay could provide a window of opportunity to study pre-treatment CAR.

Third, children should be serially followed during treatment, in order to measure changes in CAR and the circumstances under which these changes occur. Circumstances may include length of treatment; type of treatment including surgery, radiation and chemotherapy; and effects of specific drugs, drug combinations and multiple cycles of drugs. For example, we had no data on the number of dexamethasone pulses children received prior to participation, yet the effect of total dexamethasone exposure or number of exposures could be important determinants of outcome.

Fourth, long-term outcomes of disrupted circadian rhythms should be measured. Outcomes should include QoL, fatigue, changes in neurocognitive and psychological status, relapse and death. We know from studies of adults with cancer that disruption of circadian rhythms can have dire consequences such as lowered responsiveness to treatment, shortened time to relapse and decreased survival [4,41], and that disruption of circadian rhythms can significantly decrease QoL [4,5]. Our study identified changes in CAR across 10 consecutive 24-hour periods, including time before and after initiation of dexamethasone. It remains unknown how long it takes for CAR to return to baseline between pulses of dexamethasone, whether CAR deteriorates across repeated courses, or whether and how long it takes for CAR to fully recover after completion of therapy.

Fifth, while research on the effects of cancer and its treatment on circadian rhythms in children proceeds, effective approaches to support and improve these rhythms should be tested in randomized controlled trials. There is research in women with breast cancer demonstrating the effectiveness of light therapy in preventing worsening of overall fatigue [42] and deterioration of CAR and QoL during chemotherapy [43,44]. These and other interventions could be tested in children, to determine their effectiveness in supporting CAR during therapy.

Studies of children with cancer present significant challenges. Among three active pediatric treatment centers over five years, the parent study from which our sample was drawn recruited 100 children with ALL. This is large for a pediatric cancer study, yet is too small to provide an adequately sized sample to determine differences in subgroups such as adolescents. Discovering answers to important questions, including those we have posed, will require recruitment across multiple sites, and adequate funding to coordinate and carry out large-scale studies.

In conclusion, our findings suggest that dexamethasone therapy dampens CAR in children with ALL on continuation chemotherapy, and that deterioration of CAR is associated with increased fatigue. We suggest a series of future research priorities to investigate cancer-related changes in CAR, long-term outcomes of these changes and measures that might effectively minimize disruption of CAR in children treated for cancer.