PHYSICAL FITNESS AND NEUROCOGNITVE OUTCOMES IN ADULT SURVIVORS OF CHILDHOOD ACUTE LYMPHOBLASTIC LEUKEMIA: A REPORT FROM THE ST. JUDE LIFETIME COHORT.

Abstract

Survivors of childhood acute lymphoblastic leukemia (ALL) are at increased risk for both treatment-related exercise intolerance and neurocognitive deficits. This analysis aimed to identify the association between exercise intolerance and neurocognitive impairments in ALL survivors. Cardiopulmonary exercise testing, results from a two-hour standardized neuropsychological assessment, and self-report questionnaires were obtained on 341 adult survivors of childhood ALL and 288 controls. Multivariable modeling tested associations between oxygen uptake at 85% estimated heart rate (rpkVO₂) and neuropsychological test and self-reported questionnaire domains, adjusted for sex, age at diagnosis, cranial radiation, anthracycline, and methotrexate exposure and tobacco smoking status. Compared to controls, survivors had worse rpkVO₂ and performance on verbal intelligence, focused attention, verbal fluency, working memory, dominant/non-dominant motor speed, visual -motor speed, memory span, reading and math measures (all p’s < 0.001). In adjusted models, exercise intolerance was associated with decreases in performance of verbal ability, focused attention, verbal fluency, working memory, dominant motor speed, non-dominant motor speed, visual-motor speed, memory span, reading academics and math academics in survivors. This study demonstrated an association between exercise intolerance and neurocognitive outcomes. Research is needed to determine if interventions that improve exercise tolerance impact neurocognitive function in ALL survivors.

Precis:

Previous studies in non-cancer populations of children and older adults support that physical fitness interventions can have a positive impact on executive function and academic performance. This study provides evidence to support that, even with the neurotoxic effects of anticancer therapy, improving the physical fitness of pediatric cancer patients may have a similar beneficial long-term impact on the neurocognitive outcomes of survivors.

Introduction

Survivors of childhood acute lymphoblastic leukemia (ALL) are at increased risk for long-term treatment-related morbidities including exercise intolerance and neurocognitive problems.^{1, 2} There is a well-established association between cranial radiation and impaired intelligence, executive function, memory, and academic achievement among ALL survivors.³ However, elimination of cranial radiation from the treatment of childhood ALL has not completely mitigated the risk of neurocognitive problems. ALL survivors treated on chemotherapy-only protocols also have cognitive impairments,⁴ and although research has identified some of the direct adverse effects of chemotherapy (e.g. methotrexate, anthracyclines) on cognitive function in ALL survivors, not all of the variance in late neurotoxicity is accounted for in these models.⁵ As data evaluating neurocognitive function among survivors of childhood Hodgkin lymphoma and osteosarcoma indicates an association between comorbid chronic conditions and neurocognitive function, the potential impact of systemic function on neurocognitive function in ALL survivors should also be considered.^{6, 7}

Long-term survivors of childhood ALL demonstrate reduced exercise intolerance.⁸ Exercise intolerance is a limitation in peak exercise capacity, or the maximum ability of the cardiopulmonary system to extract oxygen from blood and deliver oxygen to skeletal muscle, and of skeletal muscle to extract oxygen from the blood. In some cases this is related to underlying chronic disease, and in other cases, simply to a sedentary lifestyle.⁹ This is important as research in non-cancer populations shows that physical fitness correlates with performance in executive function, intelligence, academic achievement, and memory tasks;¹⁰ physical activity is associated with lower incidences of dementia;¹¹ and improving exercise capacity positively impacts neurocognitive function.^{12, 13}

Although animal models indicate that physical activity has been observed to upregulate neurotrophic factors and enhance synaptogenesis, neurogenesis, and angiogenesis,¹⁰ there is a scarcity of research evaluating the association between exercise tolerance and neurocognitive function in young adult survivors of childhood ALL. Therefore, we utilized a clinically-assessed cohort of childhood cancer survivors participating in the St. Jude Lifetime Cohort (SJLIFE) study to determine if exercise intolerance, expressed as decreased oxygen uptake, is associated with neurocognitive impairments in long-term survivors of childhood ALL¹⁴ and evaluated if exercise intolerance mediates the association between chronic cardiac or pulmonary conditions and neurocognitive impairment.

Subjects and Methods

Study design and participants

Using the St. Jude Lifetime (SJLIFE) cohort, we conducted a cross-sectional study of 365 survivors (88% participation rate) of childhood ALL and 306 community controls.¹⁴ ALL survivors were treated at St. Jude Children’s Research Hospital between 1980 and 2003, had survived at least five years from diagnosis, and were 18 years of age or older at evaluation.¹⁵ To ensure a representative sample, prior to recruitment, the eligible ALL population was stratified by time from diagnosis, sex and previous exposure to cranial radiation. They were then randomized within strata and recruited in order by strata until the appropriate sample size was achieved. Survivors with active disease, currently being treated for cancer, or diagnosed with non-treatment-related neurocognitive (e.g., trisomy 21), musculoskeletal or cardiopulmonary impairment were not eligible for the current study. Community controls who met the same inclusion criteria, but who had no history of childhood cancer, were recruited from a pool of adult siblings, parents, relatives or friends of the current or former St. Jude patients. Control participants were non-first-degree relatives of SJLIFE survivor participants (Figure 1). This study was approved by the institutional review board at St. Jude Children’s Research Hospital, and all subjects provided written informed consent to participate.

Figure 1.

Consort diagram of participants in this study

Procedures

Because of the increased risk for cardiac events in ALL survivors, study participants completed submaximal (85% of predicted heart rate) cardiopulmonary exercise testing with continuous 12-lead electrocardiogram monitoring. An initial belt speed of 2 kilometer per hour (km/h) was used for the treadmill and increased by 1 km/h until the participant’s heart rate reached 85% of predicted maximum. Expired gas was collected with a two-way non-rebreather valve and monitored continuously using a breath-by-breath gas analysis system (Ultima CardiO2, MCG Diagnostics, St. Paul, Minnesota), to evaluate oxygen uptake at 85% estimated HR (relative peak volume of oxygen [rpkVO₂]). Exercise capacity was expressed as rpkVO₂ max which was the peak volume of oxygen measured at the time point in which the participants heart rate was at 85% of predicted maximal heart rate. For ease of clinical interpretation, during statistical modelling, rpkVO₂ was converted to Metabolic Equivalents (MET’s) where a single MET is defined as the amount of oxygen a person consumes (energy expended) per unit of body weight during one minute of rest. This is equal to 3.5 ml of oxygen consumption per kilogram of body weight per minute or 1 kilocalorie per kg of body weight per hour.¹⁶

Neuropsychological testing was conducted during a two-hour session in a dedicated workspace. Both the neuropsychological assessments and treadmill testing were randomly scheduled during a three- or four-day visit. If the neuropsychological assessment occurred after treadmill testing, the participant was allowed at least a one hour of recovery before beginning neuropsychological testing. The neurocognitive domains assessed included: intelligence (Wechsler Abbreviated Scale of Intelligence [WASI-II]¹⁷), academic skills (Woodcock-Johnson III Test of Achievement¹⁸), attention (Connor’s Continuous Performance Test II¹⁹, Trail Making Test [TMT] part A²⁰), memory (California Verbal Learning Test II²¹, Visual Selective Reminding²², WASI-III [digit span forward]²³), processing speed (Grooved Pegboard Test²⁴, WAIS-III [symbol search and digit symbol coding]²³), and executive function (TMT part B²⁰, Controlled Oral Word Association Test²⁵, WAIS-III [digit span backward]²³). In addition, survivors completed a self-rating questionnaire (Behavior Rating Inventory of Executive Function [Adult version]²⁶), to evaluate perceived neurobehavioral function. Order of testing was standardized, and survivors’ schedules were adjusted to limit impact from fatigue. Neurocognitive performance and self-rating questionnaire outcomes were converted into age-adjusted z-scores (mean, 0; standard deviation, 1.0) using national norms.

Treatment data including radiation therapy exposure and cumulative doses of anthracycline and methotrexate chemotherapy was obtained from medical records by trained abstractors and categorized a priori as: cranial radiation exposure (none, > 0 <24 Gy, or ≥ 24 Gy), cumulative high dose methotrexate (continuous variable), intrathecal methotrexate, and anthracycline exposure (none, >0 to < 250 mg/m², or ≥ 250 mg/m²). Smoking status was self-reported. Participants were categorized as never smoked (defined as <100 lifetime cigarettes), former smoker, or current smoker. Finally, we used a modified National Cancer Institute’s Common Terminology Criteria for Adverse Event (CTCAE version 4.03) to grade treatment related pulmonary and cardiac morbidities recorded prior to assessment²⁷. We dichotomized the chronic condition variable (low grade 0–1 vs high grade ≥ 2) based on the highest recorded grade for either condition.

Statistical Analysis

Demographic characteristics were compared between survivors and controls with two-sample t-tests or non-parametric equivalents if appropriate. Mean values of age- and sex-adjusted rpkVO₂ and mean age-adjusted Z-scores of objective and self-reported neurocognitive outcomes were calculated and compared between survivors and controls using two sample t-tests. Survivors’ neurocognitive scores were compared to the normal distribution using a one sample t-test. Using a closed testing procedure to correct for multiple comparisons, we selected those neurocognitive and patient-reported outcomes that differed from controls and population norms. Multivariable modeling was conducted to explore the association between those neurocognitive and patient-reported outcomes and rpkVO₂ (converted to METs for improved clinical relevance). Survivor models were adjusted for sex, age at assessment, cranial radiation, anthracycline and methotrexate exposure, and smoking status. Control models were adjusted for sex, age at assessment and smoking status. We conducted a comparison of median METS between survivors with or without neurocognitive impairment (z-score 2 standard deviations or more below the mean) utilizing the Wilcoxon signed-rank test. To assess if the association between the presence of a cardiac/pulmonary chronic condition and neurocognitive outcome was mediated by cardio-respiratory fitness (rpkVO₂), we used three stages of generalized linear regression models as described by Baron and Kenny²⁸ for each neurocognitive outcome. The first regressed rpkVO₂ on the presence of a chronic condition; the second regressed neurocognitive outcome on the presence of a chronic condition; and the third regressed neurocognitive outcome on rpkVO₂ and the presence of a chronic condition. The mediated effect via rpkVO₂ was calculated using estimates from the first and third regression for each neurocognitive outcome and a significance test was conducted using the t-statistic at a type one error of p<0.05. Participants with missing neurocognitive data were list wise deleted. All analyses were conducted using SAS software version 9.4 (SAS Institute, Inc., Cary, North Carolina). For original data used in this research, please contact Kiri.Ness@stjude.org.

Results

In this cross-sectional analysis, we evaluated 341 survivors of ALL (49% female; mean age at diagnosis 5.1 [range 0.6–18.8] years; mean age at evaluation 28.5 [range 18.4–44.6] years) and 288 controls (49.3% female; mean age at evaluation 32.2 [range 18.3–44.8]) years. Survivors did not differ from controls by sex and race but were younger at evaluation (median (range), 28.5 (18.4–44·6) vs. 32.2 (18.3–44.8) years, p<0.001) (Table 1 and supplement Table A). After adjusting for age and sex, survivors had lower mean relative peak VO₂ measurements (23.45±7.37 vs 33.03±7.39 milliliters per kilogram per minute (ml/kg/min), p<0.001) than controls (Table 2 and supplement Table B). There was a significant difference (p<0.001) in the median time to reach 85% predicted HR between survivors (12.33 minutes, range: 3.2–18.4) and controls (10.7 minutes, range 4.0–17.2). This could represent a blunted response to exercise, induced by autonomic dysfunction.²⁹ Survivors demonstrated lower performance on measures of intelligence, attention, executive function, processing speed, memory, and academics compared to controls and expected population normal values. Survivors scored lower than controls on three measures of attention: focused attention (z-score −0.23 vs 0.51, p<0.001), commissions (z-score −0.15 vs 0.06, p=0.01) and detectability (z-score −0.13 vs 0.11, p=0.002) as well as in all executive functions, processing speed, memory, and academic tests. Based on reported outcomes (higher z-scores are worse), survivors had worse emotional control (0.18 vs −0.08, p=0.007), self-initiation (0.15 vs −0.07, p=0.01) and working memory (0.56 vs. 0.29, p<0.001) than controls.

Table 1.

Characteristics of ALL survivors and community controls.

	Survivors (N=341)		Controls (N=288)		p-value
Age, Median (Range)	28.5	(18.4,44.6)	32.2	(18.3,44.8)	<0.001a
Diagnosis age, Median (Range)	5.1	(0.6,18.8)	-	-	-
Survival time, Median (Range)	21.8	(11.0,30.7)	-	-	-
Sex
Male, N (%)	174	(51.0)	139	(48.2)	0.49b
Female, N (%)	167	(49.0)	149	(51.7)
Race
Non-White, N (%)	46	(13.5)	31	(10.8)	0.30b
White, N (%)	295	(86.5)	257	(89.2)
Smoker
Never	223	(65.40)	202	(70.14)	<0.001b
Current	82	(24.05)	34	(1181)
Past	36	(10.56)	52	(18.06)

^aIndicates based on Wilcoxon two-sample test

^bIndicates based on Chi-Square test

Table 2.

Neurocognitive testing and patient-reported outcomes among ALL survivors and community controls.

	Survivors (N=341)		Controls (N=288)		p-value	One Sample t-test (compared to normal distribution)
	Mean (SD)	%	Mean (SD)	%		t-statistic	p-value
Fitnessa
VO2 85% pred HR	23.45(7.37)	-	33.03(7.39)	-	<0.0001***	-	-
Time to 85% pred	12.3(3.2)	-	10.7(4.0)		<0.0001***
Neurocognitive testingb
Intelligence
Verbal ability	−0.35(1.10)	19.06	0.17(0.92)	5.90	<0.0001***	−5.90	<0.0001***
Perceptual ability	0.02(0.91)	8.21	0.25(0.72)	3.13	<0.0003***	0.32	0.7525
Full Scale IQ	−0.16(0.92)	10.85	0.26(0.79)	2.78	<0.0001***	−3.12	0.0019**
Attention
Focused Attention	−0.23(1.33)	15.84	0.51(0.90)	3.82	<0.0001***	−3.23	0.0015**
Omissions	−0.02(1.23)	9.38	0.18(0.93)	5.56	0.03*	−0.30	0.7674
Commissions	−0.15(1.06)	14.08	0.06(1.02)	9.38	0.0105**	−2.69	0.0075**
Variability	−0.04(1.03)	10.56	−0.02(1.09)	10.07	0.79	−0.72	0.4731
Detectability	−0.13(0.92)	6.45	0.11(0.93)	3.82	0.002**	−2.63	0.0089**
Response style	0.16(0.69)	3.23	0.18(0.66)	2.78	0.59	4.31	<0.0001***
Executive
Cognitive Flexibility	−0.66(1.67)	24.93	0.27(1.13)	7.64	<0.0001***	−7.34	<0.0001***
Verbal fluency	−0.24(1.07)	19.65	0.11(1.06)	11.46	<0.0001***	−4.18	<0.0001***
Working Memory	−0.26(0.90)	7.92	0.04(0.87)	3.13	<0.0001***	−5.34	<0.0001***
Processing Speed
Dominant motor speed	−0.92(1.31)	30.21	−0.11(1.09)	12.85	<0.0001***	−12.90	<0.0001***
Non-dominant motor speed	−0.81(1.38)	27.57	−0.04(0.91)	9.38	<0.0001***	−10.87	<0.0001***
Visual-motor speed	−0.22(0.96)	15.84	0.43(0.95)	4.17	<0.0001***	−4.28	<0.0001***
Visual speed	0.07(0.97)	8.80	0.63(0.95)	2.08	<0.0001***	1.37	0.1702
Reaction time	−0.11(0.95)	9.97	0.18(1.03)	7.99	0.0003**	−2.18	0.0300*
Memory
Memory Span	−0.16(1.03)	9.97	0.18(0.94)	4.17	<0.0001***	−2.81	0.0053**
New learning	0.01(1.16)	10.26	0.33(1.03)	7.99	0.0004***	0.15	0.8847
Academics
Reading	−0.30(0.65)	6.16	0.002(0.53)	1.39	<0.0001***	−8.48	<0.0001***
Math	−0.43(0.89)	12.61	−0.14(0.85)	8.33	<0.0001***	−8.90	<0.0001***
Patient reported outcomesc
Inhibitory control	0.11(1.00)	11.14	0.08(1.00)	11.11	0.73	2.04	0.0421
Behavioral flexibility	0.19(1.14)	16.72	0.03(1.00)	9.38	0.08	3.04	0.0025**
Emotional control	0.18(1.18)	17.30	−0.08(1.06)	9.72	0.007**	2.79	0.0056**
Self-monitoring	−0.07(1.14)	13.20	−0.22(0.98)	7.29	0.11	−1.13	0.2597
Behavioral index	0.14(1.15)	15.84	−0.05(1.00)	7.64	0.04*	2.30	0.0221*
Self-initiation	0.15(1.12)	15.84	−0.07(0.94)	8.33	0.01*	2.51	0.0127*
Working memory	0.56(1.34)	27.57	0.29(1.06)	17.01	0.0008*	7.69	<0.0001*
Planning	0.11(1.05)	14.37	0.01(1.01)	11.11	0.23	1.98	0.0488*
Task completion	0.17(1.10)	12.90	0.02(1.04)	10.07	0.09	2.83	0.0050**
Organization	0.08(1.10)	13.78	−0.06(1.11)	9.72	0.13	1.33	0.1860
Metacognition	0.25(1.14)	19.65	−0.04(1.03)	10.42	0.02*	3.99	<0.0001***
Global executive composite	0.22(1.15)	17.89	−0.01(1.03)	10.07	0.01*	3.48	0.0006**

^aIndicates age- and sex-adjusted VO₂ at 85% predicted HR is in units of ml/kg/min

^bIndicates neurocognitive mean scores are age adjusted Z-scores.

^cIndicates higher patient reported outcome mean scores indicates worse outcomes.

^*p<0.05,

^**p<0.01,

^***p<0.001

In multivariable models, adjusting for age, sex, cranial radiation, anthracyclines, high-dose methotrexate, intrathecal methotrexate, smoking status and physical activity, a one metabolic equivalent increase in exercise tolerance was associated with increases in performance of verbal ability (β=0.10, p=0.010, focused attention (β=0.096, p=0.04), verbal fluency (β=0.13, p<0.001), working memory (β=0.07, p=0.03), dominant motor speed (β=0.15, p=0.001), non-dominant motor speed (β=0.15, p=0.002), visual-motor speed (β=0.010, p=0.004), memory span (β=0.11, p=0.004), reading academics (β=0.05, p=0.03) and math academics (β=0.098, p=0.001) in survivors (Table 3 and supplemental Table C). Survivors with impaired neurocognitive scores (defined as two standard deviations below the mean) were identified in the following areas: verbal ability (9.7%), verbal fluency (6.2%), dominant motor speed (15.5%), non-dominant motor speed (15.2%), memory span (2.6%), math (5.0%), and focused attention (7.9%). Among these survivors, those with impaired verbal ability (p=0.03) and focused attention scores (p=0.03) had significantly lower median METs than non-impaired survivors (Figure 2). The results of our mediation analysis indicated that relative peakVO2 did not mediate the association between presence of a chronic condition and individual neurocognitive outcomes (data not shown).

Table 3.

Associations between neurocognitive outcomes and METS among ALL survivors. An increase in activity tolerance from sitting on the couch watching television (1 MET) to riding a stationary bicycle with light effort (5.5 MET) is associated with a clinically significant improvement in every highlighted domain.

Outcome	βa	95% CI	P
Verbal ability (intelligence)	0.10	(0.024, 0.18)	0.010
Focused attention (attention)	0.096	(0.0045, 0.19)	0.04
Commissions (attention)	0.063	(−0.0098, 0.14)	0.09
Detectability (attention)	0.051	(−0.013, 0.12)	0.12
Cognitive flexibility (executive)	0.081	(−0.035, 0.20)	0.17
Verbal fluency (executive)	0.13	(0.053, 0.20)	0.0008
Working memory (executive)	0.07	(0.0062, 0.13)	0.03
Dominant motor speed (processing speed)	0.15	(0.062, 0.24)	0.001
Non-dominant motor speed (processing speed)	0.15	(0.058, 0.25)	0.002
Visual-motor speed (processing speed)	0.010	(0.031, 0.16)	0.004
Memory span (memory)	0.11	(0.035, 0.18)	0.004
Reading (academics)	0.050	(0.0050, 0.095)	0.03
Math (academics)	0.098	(0.039, 0.16)	0.001
Emotional control (self-report)b	−0.0015	(−0.080, 0.077)	0.97
Behavioral index (self-report)b	0.0020	(−0.075, 0.079)	0.96
Self-initiation (self-report)b	−0.072	(−0.15, 0.0054)	0.068
Working memory (self-report)b	−0.034	(−0.13,0.06)	0.48
Metacognition (self-report)b	−0.028	(−0.11, 0.052)	0.49
Global executive compositeb	−0.017	(−0.096, 0.063)	0.68

Abbreviations: METS, metabolic equivalents; ALL, acute lymphoblastic leukemia.

^aIndicates, in a multivariable model adjusted for age, sex, cranial radiation, anthracyclines, methotrexate and smoking status, a one MET increase in exercise tolerance is associated with the β value increase in the z-score.

^bIndicates that in patient reported outcome higher mean scores indicates worse outcomes, so β associated with worse outcomes would be negative.

METS = 85% predicted HR VO2 / 3.5

Figure 2.

Comparison of median METS and age adjusted neurocognitive z-scores between impaired and non-impaired survivors.

Discussion

This study demonstrates an independent association between exercise intolerance and lower scores on measures of intelligence, attention, executive functions, processing speed, memory, and academic performance in young adult survivors of childhood ALL. In addition, both exercise intolerance, measured by relative peakVO₂ during submaximal cardiopulmonary exercise testing, and neurocognitive performance on measures of intelligence, attention, executive functions, processing speed, memory, and academics are significantly lower in survivors of childhood ALL than among slightly older community controls. Our mediation analysis indicated that while having a chronic condition impacted exercise intolerance, exercise intolerance as a result of a chronic cardiac or pulmonary condition explained some, but not all of the association between chronic disease and impaired cognition. Therefore, exercise intolerance did not completely mediate the relationship between the presence of a chronic cardiac or pulmonary condition and neurocognitive outcomes in our sample.

One possible explanation for this finding is that, even in the presence of chronic disease, exercise has a direct effect on brain function by inducing the expression of brain-derived neurotropic factor and nerve growth factors.³⁰ Synaptic plasticity, which is important for learning and memory, is regulated by the actions of brain-derived neurotropic factor. Exercise also mitigates the harmful consequences of stress.³¹ For example, rats exposed to chronic wheel running had alterations in the serotonergic and norepinephrine systems which attenuated expression of stress-induced cortical β-adrenoceptors, hence mitigating the exaggerated response seen in uncontrolled stress.³² Additionally, the literature implicates the protective effects of vascular endothelial growth factor and insulin-like growth factor 1 as additional cellular responses to physical activity.³³ In our study, we found significant associations between and exercise intolerance and neurocognitive performance across most but not all neurocognitive domains. This may indicate that there is a protective effect of global network activity related to some tasks, and or that injury to core networks may produce significant declines in multiple domains. This is supported by data from Stevens, et al., who identified five distinct networks associated with errors of commission scores, and Garrison et al., who reported that difficulty with focused attention tasks is associated with disengagement of the core default mode network.^{34, 35}

As previous data from this cohort indicates that exercise intolerance is affected primarily by inactivity. Thus, interventions to increase physical activity have the potential to remediate neurocognitive loss among survivors of childhood ALL.¹⁵ Because previous studies report reduced levels of activity soon after completion of therapy among survivors of childhood ALL,³⁶ and because exercise intolerance and neurocognitive impairment begins early in this population,³⁷ early intervention may be key in reducing neurocognitive decline in survivors. Evidence for this is demonstrated in a meta-analysis of studies investigating physical activity level and neurocognitive performance among healthy, mentally impaired or physically disabled school age children (4–18 years old).³⁸ There is a reported positive association between physical activity and performance on measures of perception, intelligence quotients, achievement, verbal abilities, mathematical skill, and academic readiness. Additionally, Raine, et al. demonstrated that positive changes in physical fitness (or exercise tolerance) among adolescents have significant benefit in academic achievement. The authors found that obese children who showed improvements in aerobic fitness (as measured by 15 meter shuttle laps with increasing pace until participant failure) between sixth and eighth grade had greater cognitive gains (determined by Idaho Standards Achievement Test reading and math scores) and academic achievement than healthy weight age matched controls, independent of changes in whole-body fat.³⁹ In another study, Jackson et al. found a significant association between self-reported physical activity and verbal ability (a measure of intelligence).⁴⁰ Specifically, a same-sex twin who was more sedentary during adolescence than their sibling was found to have lower verbal ability during early adulthood even after controlling for environmental and genetic factors. In our study, a small increase in exercise tolerance (6 MET = stationary bicycling with light effort vs. 7 MET = jogging or light calisthenics) was associated with a significant increase in verbal ability and focused attention (a measure of attention).

There is evidence that young ALL survivors can benefit from structured physical activity training. In one study, seventeen 16- to 30-year old survivors were asked to participate in a 16-week low-cost home-based exercise program that included muscle training and aerobic exercise.⁴¹ The home exercise program improved both the mean relative peakVO2 and maximal work load of the participants. Moreover, these benefits were not limited to those who have completed therapy. Children in the maintenance phase of chemotherapy who participated in a 6-month program of stretching, strengthening and aerobic exercise also demonstrated improvements in cardiopulmonary fitness as measured by the modified six-minute walk test.⁴² These studies demonstrate that a low-cost home-based exercise training program can effectively increase cardiopulmonary fitness during and after completion of therapy in young adult survivors of childhood ALL, and reduce the risk for neurocognitive late effects in this population.

Our findings should be considered in the context of several study limitations. First, as our evaluation was cross sectional, we could not determine the direction of the association between fitness and neurocognitive outcomes. Secondly, our healthy controls were recruited from families and friends of current or former St. Jude patients treated for a variety of cancers. As such they may not represent the population as a whole but are more representative of the social and economic demographics of the survivors treated at our institution. It is possible that our healthy controls had above average fitness which may have inflated our estimates of the differences between survivors and members of the general population. However, this potential bias is unlikely because the relative peakVO₂ levels of our controls are lower than those of other healthy populations in similarly published studies.⁴³ Additionally, our assessment of exercise capacity was from a submaximal test (85% of predicted heart rate). Given that there is high variability in the relative workload that each participant can perform at 85% of maximal heart rate, this may have introduced some measurement error, limiting the precision of our estimates. Although some survivors may have experienced prolonged hospitalization that disrupted the normal educational experience, previous studies show that when compared to healthy same age peers and to other hospitalized pediatric patients with diseases that do not include central nervous system involvement, ALL survivors demonstrate poorer academic achievement and neurocognitive deficits.⁴⁴ Finally, this study was conducted on a survivor population treated at a single institution, which may impact its generalizability to survivors treated elsewhere.

In conclusion, this study links exercise intolerance with neurocognitive impairment in adult survivors of childhood ALL, a finding consistent with previous studies reporting associations between exercise intolerance and cognitive decline in healthy older adults.⁴⁵ The associations between exercise intolerance and lower scores on measures of intelligence, executive function, processing speed, memory, and academics suggest that improving the exercise tolerance among children with ALL, after treatment, may benefit neurocognitive and academic performance of the survivors. A well-designed randomized prospective study is needed to evaluate if enhancing exercise tolerance would improve neurocognitive outcomes among childhood ALL survivors.