Premature aging as an accumulation of deficits in young adult survivors of pediatric cancer

Abstract

Background

We aimed to characterize premature aging as an accumulation of deficits in survivors of pediatric cancer compared with community controls and examine associations with host and treatment factors, neurocognition, and mortality.

Methods

Pediatric cancer survivors (n = 4000, median age = 28.6, interquartile range [IQR] = 23-35 years; 20 years postdiagnosis: IQR = 15-27), and community participants without a history of cancer serving as controls (n = 638, median age = 32, IQR = 25-40 years) completed clinical assessments and questionnaires and were followed for mortality through April 30, 2020 (mean [SD] follow-up = 7.0 [3.4] years). A deficit accumulation index (DAI) score was calculated from 44 aging-related items including self-reported daily function, psychosocial symptoms, and health conditions. Items were weighted from 0 (absent) to 1 (present and/or most severe), summed and divided by the total yielding a ratio (higher = more deficits). Scores less than 0.20 are robust, and 0.06 is a clinically meaningful difference. Linear regression compared the DAI in survivors and controls with an age*survivor or control interaction. Logistic regression and Cox-proportional hazards estimated the risk of neurocognitive impairment and death. Models were minimally adjusted for age, sex, and race and ethnicity.

Results

The adjusted mean DAI among survivors at age 30 years was 0.16 corresponding to age 63 years in controls (33 years premature aging; β = 0.07, 95% confidence interval [CI] = 0.06 to 0.08; P < .001). Cranial and abdominal radiation, alkylators, platinum, and neurosurgery were associated with worse DAI (P ≤ .001). Higher scores were associated with increased risk of neurocognitive impairment in all domains (P < .001) and increased risk of death (DAI = 0.20-0.35, hazard ratio = 2.80, 95% CI = 1.97 to 3.98; DAI ≥ 0.35, hazard ratio = 5.08, 95% CI = 3.52 to 7.34).

Conclusion

Pediatric cancer survivors experience clinically significant premature aging. The DAI may be used to identify survivors at greatest risk of poor health outcomes.

Aging has been described as an accumulation of molecular damage ultimately leading to tissue and organ dysfunction and physiologic dysregulation (1). In the general population, age is the most significant risk factor for outcomes such as cancer, functional decline, neurocognitive impairment, and mortality. The substantial variability in development of disease and functional decline among individuals of the same chronological age suggests heterogeneity in age-related accumulation of damage.

This variability may be exacerbated in cancer survivors whose treatment causes molecular damage, including changes to DNA structure, cellular function, signaling, and tissue integrity (2). In pediatric patients, these exposures happen during development and may set survivors on a unique aging trajectory. Further, early molecular damage may make survivors more vulnerable to the negative health implications of social and environmental influences encountered during aging (eg, socioeconomic factors, smoking) (3). Survivors of pediatric cancer have higher than expected rates of chronic health conditions, physiologic frailty, and neurocognitive impairment as well as elevated telomere attrition and epigenetic age, all of which suggest premature aging (4-11).

We propose a deficit accumulation index (DAI) framework to assess aging, which will expand previous work in survivors of pediatric cancer by considering the accumulation of multiple deficits across various health and functional domains (12). A comprehensive and integrated approach to measuring individual aging is important because components that contribute to aging are multifactorial and interdependent. Accumulation of multiple and interdependent deficits increases the risk of mortality and, theoretically, biologic aging and physiologic dysregulation (12,13). Using this approach, integration of large and small effects can be considered using information that on its own may only weakly correlate with an outcome. For example, the accumulation of deficits has predicted dementia in the general population when individual risk factors could not (14).

We aimed to characterize aging as an accumulation of deficits in survivors of pediatric cancer relative to community controls, identify associated host and treatment factors, and examine associations with aging-related outcomes including neurocognition and mortality. The DAI framework has not been applied in survivors of pediatric cancer and provides multiple advantages over previous work done in this population (eg, chronic health cumulative burden, frailty). Including a focus on assessment of diverse aging-related items across multiple systems and the DAI may be used on a continuous scale to detect more subtle, but clinically meaningful, changes over time that may help assess response in intervention trials. Importantly, administering the DAI requires no specific training. Aging-related items included in the DAI may be extracted from medical records or assessed by questionnaires making them a feasible way to identify survivors at risk for aging-related outcomes.

Methods

Participants

The St. Jude Lifetime Cohort (SJLIFE) was established to facilitate prospective clinical assessment of outcomes among survivors of pediatric cancers (15). Eligible survivors were diagnosed between 1962 and 2012, treated at St. Jude Children’s Research Hospital, survived at least 5 years from diagnosis, and were aged 18 years or older. Particiants from the community without a history of cancer served as controls. Of 4392 eligible survivors and 718 controls, 4000 (91%) and 683 (95%), respectively, were included in these analyses (Supplementary Figure 1, available online). St. Jude Children’s Research Hospital institutional review board approved this protocol; all participants provided informed consent.

Procedures

Following the methods of Rockwood and Searle (16), we designed a DAI using 44 items from SJLIFE questionnaires, clinical examinations, and medical records (Supplementary Table 1, available online) (17). DAIs are validated to predict aging-related outcomes like hospitalization, mortality, and neurocognitive decline in noncancer populations and adult cancer patients (12-14,16,18-20). A DAI can include various items, provided they cover a range of comorbidities, physical, and psychosocial functions related to health and increase in prevalence with age. The number and specific deficits do not need to be the same across studies to replicate findings, as long as at least 30 items are included (16). Participants missing more than 10% of items were excluded (n = 392) (16). Each item is weighted from 0 to 1, where 1 indicates the item is present and/or most severe. Item weights are summed and divided by the number of items available, resulting in a ratio (range 0 to 1 higher worse). A large, clinically meaningful difference is 0.06 (21). Our primary DAI measure was continuous. To increase interpretability, we also measured DAI in categories identified in previous research as associated with hospitalization and mortality (low <0.2; medium 0.2 to <0.35, high ≥0.35) (18,22-24).

Survivors were followed annually to ascertain vital status (through April 30, 2020). Neuropsychological tests were conducted concurrently with the clinical assessment and questionnaires. Domains assessed include attention (25,26), processing speed (27,28), memory (29), and executive function (Supplementary Table 2, available online) (26,28,29). Scores were referenced to normative data to generate age-adjusted z scores. Neurocognitive impairment was defined as a z score less than the 10th percentile (30,31). Neurocognitive analyses excluded 26 participants without a neuropsychological exam and 136 participants who were non-English speaking or had a genetic, neurodevelopmental syndrome, or neurologic injury (eg, traumatic brain injury).

Cancer type and treatment history included chemotherapy, surgical procedures, and region/organ-specific radiation dosimetry based on medical records. Other covariates included self-reported sociodemographics (eg, race [American Indian or Alaskan Native, Asian, Black, Pacific Islander, Other, White], ethnicity [Hispanic, non-Hispanic, Other], marital status, education, employment, and income), health behaviors (smoking, sedentary behavior: 2 or more hours screen time outside work), and alcohol use. Race and ethnicity were combined to form non-Hispanic Black, non-Hispanic White, and other categories; other included American Indian, Alaskan Native, Asian, Pacific Islander, Hispanic Black, Hispanic White, multiple race, and those who reported Caribbean, Cuban, Mexican or Chicano, Puerto Rican, or South or Central American ethnicity.

Statistical analysis

Survivors and controls were compared on covariates using χ² or 2-sided t tests. Linear regression models estimated differences in mean DAI in survivors (overall and by diagnosis) and controls adjusting for current age, sex, and race and ethnicity. Unadjusted locally estimated scatterplot smoothing (LOESS) curves confirmed the linear relationship across the age range of the sample (18 to 70 years). We included an age by survivor or control interaction term to test if survivors accumulated deficits faster than controls. Multinomial logistic regression estimated the probability of being in each of the 3 categories of DAI (low, medium, high) in survivors compared with controls adjusting for the same covariates.

Univariate linear regression models estimated associations between mean DAI (continuous) with each cancer treatment (vs absence of treatment). Statistically significant treatment associations (P < .05) were combined in a multivariable model adjusted for current age, sex, and race and ethnicity. Multinomial logistic regression estimated the probability of medium or high DAI scores vs low scores associated with each treatment. Demographic, socioeconomic, and behavioral factors were examined similarly.

Among survivors, modified Poisson regression models estimated the risk of neurocognitive impairment associated with the medium and high vs the low DAI group. All neurocognitive models were adjusted for current age, age at diagnosis, sex, race and ethnicity, and Children’s Oncology Group (COG)–identified risk factors for neurocognitive impairment (ie, neurosurgery, cranial radiation, intrathecal methotrexate, high-dose intravenous methotrexate or cytarabine) (32). Sensitivity analyses were conducted excluding survivors who received any COG-identified risk factors for neurocognitive impairment (n = 2625 excluded) (33). Among survivors, Cox proportional hazards models estimated the risk of all-cause mortality associated with being in the medium or high vs low DAI groups adjusted for current age, sex, race and ethnicity, and year of diagnosis.

The distribution of the DAI was slightly skewed, however, results from analyses after log transformation were similar to untransformed models. Therefore, results from untransformed models are presented. All statistical tests were 2-sided, and a P value less than .05 was considered statistically significant. Analyses were conducted in SAS 9.4 (SAS Institute, Cary, NC, USA).

Results

Survivors were a median age of 28 (interquartile range [IQR] = 23.2-35.5 ) years and 20 years postdiagnosis (IQR = 14.6-27.4) at evaluation. Community controls were slightly older and more likely to be female (Table 1). Survivors were more likely to be overweight or obese; have a lower educational and employment attainment and lower income; and be a current smoker.

Table 1.

Clinical and demographic characteristics of survivors and community controls at baseline entry into the St. Jude Lifetime cohort

Characteristic	Survivor	Community control
	(n = 4000)	(n = 683)
Median age at evaluation (IQR), y	28.6 (23.2-35.5)	31.5 (24.7-39.9)
Sex, No. (%)
Female	1901 (47.5)	385 (56.4)
Male	2099 (52.5)	298 (43.6)
Race and ethnicity, No. (%)a
Non-Hispanic Black	574 (14.4)	44 (6.5)
Non-Hispanic White	3224 (80.6)	551 (80.9)
Other	202 (5.1)	86 (12.6)
BMI, No. (%)
Normal	2498 (62.5)	457 (66.9)
Underweight	145 (3.6)	15 (2.2)
Overweight or obese	1357 (33.9)	211 (30.9)
Education, No. (%)b
<High school	353 (9.6)	18 (2.8)
High school or GED	769 (21.0)	75 (11.6)
Some training or college	1333 (36.4)	193 (29.8)
College degree	906 (24.7)	235 (36.3)
Postgraduate degree	302 (8.2)	126 (19.5)
Employment, No. (%)c
Unemployed	1086 (27.9)	78 (11.7)
Part-time	596 (15.3)	99 (14.8)
Full-time	2205 (56.7)	491 (73.5)
Marital status, No. (%)d
Single or never married	1804 (46.8)	226 (33.8)
Widowed	14 (0.4)	3 (0.4)
Divorced or separated	401 (10.4)	52 (7.8)
Married or living as married	1633 (42.4)	388 (58.0)
Household income, No. (%)e
≤$19 999	660 (19.8)	58 (9.3)
$20 000-$59 999	1403 (42.0)	219 (35.0)
≥$60 000	1277 (38.2)	348 (55.7)
Smoking status, No. (%)f
Former smoker	390 (9.8)	86 (12.6)
Current smoker	763 (19.2)	99 (14.5)
Never smoker	2816 (71.0)	496 (72.8)
Median age at diagnosis (IQR), y	8.1 (3.5-13.6)	—
Median time since diagnosis (IQR), y	20.0 (14.6-27.4)	—
Diagnosis, No. (%)
Acute lymphoblastic leukemia	1199 (30.0)	—
Acute myeloid leukemia	171 (4.3)	—
CNS tumor	540 (13.5)	—
Ewing sarcoma	115 (2.9)	—
Hodgkin lymphoma	464 (11.6)	—
Non-Hodgkin lymphoma	277 (6.9)	—
Neuroblastoma	168 (4.2)	—
Osteosarcoma	150 (3.8)	—
Retinoblastoma	124 (3.1)	—
Soft tissue sarcoma	258 (6.5)	—
Wilms tumor	233 (5.8)	—
Others	301 (7.5)	—
Radiation, yes, No. (%)
Cranial radiationg	1148 (29.4)	—
Chest radiationh	929 (23.8)	—
Abdomen or pelvic radiationh	1173 (30.0)	—
Chemotherapy, yes, No. (%)
High-dose IV cytarabine	354 (8.9)	—
Standard-dose IV cytarabine	1265 (31.6)	—
High-dose IV methotrexate	1094 (27.4)	—
Standard-dose IV methotrexate	1024 (25.6)	—
Intrathecal methotrexate	1525 (38.1)	—
IV vincristine	2653 (66.3)	—
Anthracyclines	2335 (58.4)	—
Alkylating agent	2395 (59.9)	—
Platinum agent	597 (14.9)	—
Corticosteroids	1801 (45.0)	—
Neurosurgery, yes, No. (%)	528 (13.2)	—

^a2 missing race or ethnicity; other included American Indian, Alaskan Native, Asian, Pacific Islander, Hispanic Black, Hispanic White, multiple race, and those who reported Caribbean, Cuban, Mexican or Chicano, Puerto Rican, or South or Central American ethnicity. BMI = body mass index; CNS = central nervous system; GED = general education development; IQR = interquartile range; IV = intravenous.

^b373 missing education.

^c128 missing employment status.

^d162 missing marital status.

^e718 missing household income.

^f33 missing smoking status.

^g95 missing cranial radiation dosimetry.

^h91 missing chest or abdominal radiation dosimetry.

Deficit accumulation

Survivors had a statistically significant and clinically meaningful higher mean DAI after adjustment for age, sex, and race and ethnicity compared with controls (mean difference = 0.07, 95% confidence interval [CI] = 0.06 to 0.08; Table 2). Survivors also had an increased risk of a medium or high DAI. When plotted against age, the adjusted mean DAI at the average age of survivors (30 years) was 0.16, which corresponded to 63 years of age in controls, suggesting premature aging of 33 years (Figure 1, A). The rate of accumulation of deficits was greater in survivors compared with controls consistent with an accelerated aging phenotype (Figure 1, A; P < .001).

Table 2.

Mean deficit accumulation index (DAI), mean difference in DAI, and the relative risk (RR) of having a medium or high DAI in survivors compared with controls adjusted for age, sex, and race and ethnicity^a

	Adjusted mean DAI (95% CI)	Mean difference in DAI (95% CI)	P	Medium DAI (≥0.20 to <0.35)			High DAI (≥0.35)
				No. (%)	RR (95% CI)	P	No. (%)	RR (95% CI)	P
Survivors	0.16 (0.15 to 0.17)	0.07 (0.06 to 0.08)	<.001	792 (19.8)	4.85 (3.51 to 6.70)	<.001	336 (8.4)	8.36 (4.69 to 14.89)	<.001
Controls	0.09 (0.08 to 0.10)	0.0 (referent)		45 (6.6)	1.00 (referent)		13 (1.9)	1.00 (referent)

^aCI = confidence interval.

Figure 1.

Survivors experience accelerated deficit accumulation, consistent with the accelerated aging phenotype. A) The mean DAI (95% CI) in survivors and controls by age at evaluation adjusted for sex and race and ethnicity. The left vertical dotted line represents the mean age of survivors (30.2 years) where the adjusted mean DAI was 0.159, which was equivalent to the mean DAI of controls at age 63.8 years (right vertical dotted line). B) The adjusted mean DAI by select diagnoses and controls by age adjusted for sex and race and ethnicity. Other includes all other diagnoses. An asterisk (*) indicates slope is statistically significantly greater than the slope of the community controls (P < .05). Graphs represent entire sample age range: 18-70.2 years. ALL = acute lymphoblastic leukemia; AML = acute myeloid leukemia; CNS = central nervous system malignancy; DAI = deficit accumulation index; HL = Hodgkin lymphoma; OS = osteosarcoma.

Factors associated with DAI

When plotted against age (Figure 1, B), the rate of deficit accumulation was greater in acute lymphoblastic leukemia (P < .001) and Hodgkin lymphoma (P < .001) survivors compared with controls. The adjusted mean DAI was elevated compared with controls across all diagnostic groups and was highest in central nervous system (CNS), osteosarcoma, and acute myeloid leukemia survivors (P < .05; Figure 2, A). In multivariable models, cranial radiation, abdominal radiation, alkylators, platinum agents, and neurosurgery were associated with a statistically significantly higher DAI (P < .05; Figure 2, B; Supplementary Table 3, available online). Female sex, less than a college education, non-full-time employment, and less than $60 000 annual household income were associated with a higher DAI (P < .05; in Supplementary Table 4, available online). Current smoking (mean DAI = 0.18, 95% CI = 0.17 to 0.19) vs nonsmokers (DAI = 0.16, 95% CI = 0.15 to 0.17; P < .001) and sedentary behavior (>2 hours of sedentary behavior mean DAI = 0.18, 95% CI = 0.17 to 0.19, vs ≤2 hours DAI = 0.16, 95% CI = 0.15 to 0.17; P < .001) were also associated with a higher DAI.

Figure 2.

Adjusted mean deficit accumulation index (DAI) by diagnosis and treatment exposures. A) Adjusted mean DAI in controls and by diagnosis (adjusted for age, sex, and race and ethnicity). All diagnosis groups were statistically significantly higher than controls (P < .05). B) Adjusted mean DAI for specific treatments vs not receiving that treatment. Results are from a multivariate model mutually adjusted for treatments (including chest radiation, anthracyclines, vincristine, methotrexate, cytarabine, and corticosteroids not shown here), current age, sex, and race and ethnicity. Error bars represent 95% confidence intervals. ALL = acute lymphoblastic leukemia; AML = acute myeloid leukemia; CNS = central nervous system; HL = Hodgkin lymphoma; NHL = non-Hodgkin lymphoma; ST = soft tissue.

Neurocognition and mortality

Survivors with a medium or high DAI had a higher risk of neurocognitive impairment in all domains and outcomes compared with those with a low DAI (Table 3). A dose-response relationship between medium and high categories was observed across all neurocognitive outcomes. For example, survivors with a medium DAI had 1.6 times higher risk of impairment in short-term verbal recall (relative risk [RR] = 1.59, 95% CI = 1.34 to 1.89; P < .001), and survivors with a high DAI had 2.3 times higher risk (RR = 2.31, 95% CI = 1.90 to 2.82; P < .001) compared with those with a low DAI. Almost all associations remained statistically significant in sensitivity analyses omitting survivors treated with any COG-identified risk factors for neurocognitive impairment (Supplementary Table 5, available online).

Table 3.

The risk of neurocognitive impairment associated with a medium or high deficit accumulation index score (DAI) relative to a low DAI

Neurocognitive outcomeb	Medium vs low DAIa		High vs low DAIa
	RR (95% CI)	P c	RR (95% CI)	P c
Global cognition
Verbal reasoning	1.71 (1.47 to 1.98)	<.001	2.34 (1.98 to 2.78)	<.001
Nonverbal reasoning	1.73 (1.39 to 2.14)	<.001	2.88 (2.26 to 3.66)	<.001
Academics
Word reading	1.78 (1.37 to 2.31)	<.001	2.80 (2.09 to 3.74)	<.001
Mathematics	1.56 (1.33 to 1.83)	<.001	2.38 (1.97 to 2.87)	<.001
Attention
Sustained	1.73 (1.39 to 2.14)	<.001	3.17 (2.52 to 3.97)	<.001
Variability	1.50 (1.24 to 1.81)	<.001	2.79 (2.29 to 3.40)	<.001
Commissions	1.21 (0.98 to 1.50)	.07	2.01 (1.58 to 2.54)	<.001
Focused	1.87 (1.57 to 2.23)	<.001	2.53 (2.07 to 3.09)	<.001
Processing speed
Visual-motor	2.08 (1.79 to 2.41)	<.001	2.88 (2.44 to 3.40)	<.001
Motor	1.61 (1.45 to 1.80)	<.001	2.22 (1.94 to 2.54)	<.001
Memory
Span	1.29 (1.06 to 1.57)	.01	1.83 (1.41 to 2.37)	<.001
New verbal encoding	1.80 (1.53 to 2.13)	<.001	2.49 (2.05 to 3.03)	<.001
Short-term verbal recall	1.59 (1.34 to 1.89)	<.001	2.31 (1.90 to 2.82)	<.001
Long-term verbal recall	1.66 (1.44 to 1.92)	<.001	1.95 (1.63 to 2.33)	<.001
Executive function
Perseverations	1.40 (1.16 to 1.69)	.001	2.13 (1.73 to 2.63)	<.001
Working memory	1.47 (1.17 to 1.85)	.001	1.78 (1.28 to 2.48)	.001
Cognitive flexibility	1.43 (1.26 to 1.62)	<.001	2.09 (1.81 to 2.42)	<.001
Verbal fluency	1.40 (1.21 to 1.62)	<.001	1.68 (1.41 to 2.00)	<.001

^aThe DAI was treated as a 3-level variable with low (<0.2) as the referent group (medium: ≥0.2 to <0.35; high: ≥0.35). CI = confidence interval; RR = relative risk.

^bNeurocognitive impairment was defined as an age-adjusted z score below the 10th percentile of population norms.

^c P values reported are adjusted for multiple comparisons using the false discovery rate.

Survivors were followed for a mean (SD) of 7.0 (3.4) years from SJLIFE enrollment to death or last follow-up. Compared with survivors with a low DAI score, survivors with a medium DAI had a 2.8-fold higher risk of death (hazard ratio [HR] = 2.80, 95% CI = 1.97 to 3.98; P < .001), whereas survivors with a high DAI had a fivefold higher risk of death (HR = 5.08, 95% CI = 3.52 to 7.34; P < .001).

Discussion

Survivors of pediatric cancer are at increased risk of physiologic frailty, chronic health conditions, neurocognitive impairment, and premature mortality, all suggestive of an accelerated aging phenotype (4-8). This study expands on these previous findings, demonstrating a comprehensive accumulation of deficits, suggesting multisystem dysregulation. Our results suggest pediatric cancer and its therapies increase the accumulation of damage and disease among survivors beyond that expected based on chronological age, regardless of diagnosis. We demonstrate that survivors of pediatric cancer experience a dramatically elevated accumulation of deficits equivalent to 33 excess years of age. Notably, higher deficits accumulation was associated with neurocognitive impairment and all-cause mortality. Overall, our results support the hypothesis that cancer and its therapies, especially when experienced at young ages, are associated with premature aging.

Similar to our findings, the DAI predicts hospitalization, dementia, and mortality in the general population and cancer-related outcomes among adult cancer survivors (18,20,24). Our study extends previous research on aging-related outcomes in pediatric cancer survivors by using the accumulation of deficits rather than examining individual interdependent contributors. Consistent with the DAI, previous research in survivors of childhood cancer using the Fried frailty criteria or chronic health cumulative burden have demonstrated similar patterns across diagnoses and treatments and with mortality (5,6). In contrast, the DAI assesses a diverse set of aging-related items across multiple systems, including chronic health conditions, functional, psychosocial, and mental well-being. Therefore, our findings suggest broad multisystem dysregulation. This comprehensive assessment is important as individual aspects of aging are interdependent and the heterogeneity of DAIs may identify multiple potentially modifiable deficits. In addition, in this younger population, one important deficit may not predict adverse outcomes, but accumulation of subtle deficits taken together may increase our power to do so, potentially allowing more time for intervention (14,22). Lastly, the DAI allows us to examine relationships on a continuous scale (vs Fried frailty criteria: robust, prefrail, frail) to detect more subtle, but clinically meaningful, changes related to aging over time, which may help assess response in intervention trials.

We have primarily employed the DAI as a research tool to understand premature aging, however, a shorter version of the DAI may be a feasible method to identify survivors for risk-adapted care or early intervention to improve aging-related outcomes (34). Our DAI included systematic clinical assessments, however, most researchers and clinicians will not have access to such measures. It is important to note the items included in a DAI may be obtained via questionnaires and/or medical record abstraction, and our specific DAI items do not need to be used. Rather, our data demonstrate the methods described above can be applied to design a DAI based on available data. In fact, the DAI has been successfully adapted for automated use in the electronic medical record, effectively identifying adult cancer patients at risk for mortality (35,36). Future research is needed to identify barriers to and strategies for successful implementation of the DAI in survivorship clinics. This includes measurement studies to decrease the items and domains assessed in a similar way geriatric assessments were developed. Similar to this DAI, geriatric assessments cover multiple domains including functional status, physical performance, fall risk, comorbidities, depression, social support, nutrition, and cognition (37). These domains were derived from the DAI and selected because of clinical translatability. Geriatric assessments are feasibly administered especially when completed by the patient and electronically (38,39). The American Society of Clinical Oncology recommends geriatric assessments be standard of care among patients aged 65 years and older, to identify vulnerabilities not otherwise captured in oncologic care (37). The DAI may be a starting point for such an assessment tool for young adult survivors of cancer to aid in long-term survivorship care. We are designing a study to replicate these findings using only self-reported data and aim to narrow down the items to increase the DAI’s clinical utility and decrease clinician burden. Ideally, research would result in something similar to the Childhood Cancer Survivor Study cardiovascular and breast cancer risk prediction calculators (40,41). Additional research will be needed to clinically validate such a measure at the individual survivor level.

Consistent with higher rates of late morbidity and mortality in acute lymphocytic leukemia and Hodgkin lymphoma survivors (8,42-44), these groups accumulated deficits statistically significantly faster than community controls suggesting accelerated aging phenotypes. In contrast, CNS tumor survivors did not accrue deficits statistically significantly faster than controls, despite having the highest mean DAI. This is consistent with studies among CNS survivors demonstrating substantial morbidity that occurs early and persists over time (6). However, our data are cross-sectional, estimating change over time using chronological age. Longitudinal data are needed to confirm these patterns and examine changes by treatment era as survivors treated in modern eras, with less intensive therapy, may not accumulate deficits as quickly. Because the DAI includes many items related to functional abilities and mental health, our results support the need for extended, multifaceted survivorship care throughout the lifespan of survivors. Future research is needed on the design and implementation of survivorship care that can detect and manage the physical and mental burden of surviving pediatric cancer, which may help mitigate premature aging.

We demonstrate that higher accumulation of deficits was associated with impairment across all neurocognitive domains, including memory, executive function, and attention—a pattern consistent with dementia and Alzheimer’s disease. Consistent with our findings, the DAI is associated with impairments in attention, processing speed, executive function, and memory in breast cancer survivors (45). In the general population, DAIs predict dementia better than individual conditions or physiologic frailty (14,22). We have previously reported that chronic health conditions and physiologic frailty are associated with long-term processing speed and memory impairment (46,47). However, associations between the DAI and neurocognitive impairment appear consistent across all neurocognitive domains in a dose-response fashion. Collectively, these data support the hypothesis that physiologic and cognitive aging are tightly linked in survivors of pediatric cancers. Additional longitudinal data are needed to determine if the trajectory of aging as measured by the DAI is associated with trajectories of cognitive aging, particularly as measured by self-report as most survivors will not have access to neuropsychological testing.

We report that behaviors such as smoking and sedentary time were associated with a worse DAI. Exercise has been demonstrated to have positive effects on biologic hallmarks of aging (eg, cellular senescence, inflammation) (48) and slows the accumulation of deficits in older adult populations (49). Survivors of pediatric cancer who increased their vigorous activity over an 8-year period had a decreased risk of death (50). Emerging data suggest exercise interventions are feasible and effective in survivors of pediatric cancer (51,52) but need to be tailored to groups with limited cardiopulmonary capacity, mobility, and/or motor function (53,54). The DAI may serve as a surrogate endpoint to evaluate an intervention’s ability to slow or reverse premature aging in survivors of pediatric cancer.

This is the first study to describe premature aging in survivors of pediatric cancer using the DAI, a comprehensive measure of aging. These findings are strengthened by a well-characterized cohort of survivors of pediatric cancer with diverse diagnoses and treatment exposures. Importantly, the DAI was associated with mortality and neurocognitive functioning in a dose-dependent manner, demonstrating its relationship with aging. We plotted the DAI against chronological age to inform on the hypothesis that survivors experience accelerated aging. Although our findings support this hypothesis, results should be interpreted cautiously as our data are cross-sectional. It is possible that an age effect is conflated with a cohort effect (eg, older survivors received more intensive therapy). Additionally, although factors such as unemployment and educational attainment are associated with the DAI, it remains unclear if these factors precede the accumulation of deficits or vice versa. Our sample was limited to those aged 18 years and older, and the aging trajectory may look different in younger survivors. Future research will include adapting the DAI to those aged younger than 18 years, longitudinal data collection, and the implications of sociodemographic and behavior factors on aging among survivors of pediatric cancer.

In conclusion, survivors of pediatric cancer have a higher accumulation of deficits compared with community controls, consistent with a premature aging phenotype. Our findings highlight the need for continuous and systematic survivorship care to manage chronic health conditions and promote physical and psychosocial health. Additionally, a DAI may be a feasible assessment of aging and a tool to identify survivors most at risk for severe aging-related outcomes.

Funding

This work was supported by funding from the National Cancer Institute at the National Institutes of Health (grants K00CA222742 and K99CA256356 to AMW, KRK, and LLR, U01CA195547 to MMH and KKN, R35CA197289 to JM[J.M.], and P30CA021765 to Charlie Roberts, PhD); and from the American Lebanese Syrian Associated Charities.

Notes

Role of the funder: The funding organizations had no role in the design and conduct of the study; collection, management, analysis, and interpretation of data; preparation, review, or approval of the manuscript for publication.

Disclosures: The authors have nothing to disclose. JM, a JNCI Associate Editor and coauthor on this article, was not involved in the editorial review or decision to publish this manuscript.

Author contributions: Conceptualization: AMW, JM, BS, DS, LLR, MMH, KKN, KRK. Data curation: AMW, JM, MW, GTA, NB, TMB, WC, MJE, DAM, ZW, BS, DS, LLR, MMH, KKN, KRK. Formal Analysis: AMW, MW, DS, KRK. Funding acquisition: AMW, MMH, KKN, JM, LLR, KRK. Investigation: AMW, JM, MW, NB, TMB, WC, BS, DS, LLR, MMH, KKN, KRK. Methodology: AMW, DS, BS, JM, KRK. Project Administration: AMW, DS, KRK. Resources: AMW, MMH, KKN, JM, DS. Supervision: AMW, DS, JM, BS, KRK. Visualization: AMW, MW. Writing-Original Draft: AMW, MW, DS, KRK. Writing -review and editing: AMW, JM, MW, GTA, NB, TMB, WC, MJE, DAM, BS, ZW, DS, LLR, MMH, KKN, KRK.

Prior presentations: This work was presented, in part, at the American Society of Clinical Oncology Annual Meeting in 2021, presented virtually.

Data availability

The data underlying this article are available in the St. Jude Cloud and can be access at stjude.cloud and sjlife.stjude.org.