Cognitive Predictors of Social Adjustment in Pediatric Brain Tumor Survivors Treated with Photon versus Proton Radiation Therapy

Abstract

Background:

Pediatric brain tumor survivors are at risk for poor social outcomes. It remains unknown whether cognitive sparing with proton radiotherapy (PRT) supports better social outcomes relative to photon radiotherapy (XRT). We hypothesized that survivors treated with PRT would outperform those treated with XRT on measures of cognitive and social outcomes. Further, we hypothesized that cognitive performance would predict survivor social outcomes.

Procedure:

Survivors who underwent PRT (n=38) or XRT (n=20) participated in a neurocognitive evaluation >1 year post-radiotherapy. Group differences in cognitive and social functioning were assessed using ANCOVA. Regression analyses examined predictors of peer relations and social skills.

Results:

Age at evaluation, radiation dose, tumor diameter, and sex did not differ between groups (all p>0.05). XRT participants were younger at diagnosis (XRT M=5.0 years, PRT M=7.6 years) and further out from radiotherapy (XRT M=8.7 years, PRT M=4.6 years). The XRT group performed worse than the PRT group on measures of processing speed (p=0.01) and verbal memory (p<0.01); however, social outcomes did not differ by radiation type. The proportion of survivors with impairment in peer relations and social skills exceeded expectation (χ²(1)=38.67, p<0.001; χ²(1)=5.63, p<0.05). Household poverty predicted peer relation difficulties (t=2.18, p<0.05) and verbal memory approached significance (t=−1.99, p=0.05). Tumor diameter predicted social skills (t=−2.07, p<0.05).

Conclusions:

Regardless of radiation modality, survivors are at risk for social challenges. Deficits in verbal memory may place survivors at particular risk. Results support monitoring of cognitive and social functioning throughout survivorship as well as consideration of sociodemographic risk factors.

Introduction

Pediatric brain tumor survivors are at risk for poor long-term social outcomes relative to peers. Reviews of social outcomes in survivors have highlighted an increased risk for social skill deficits and poor long-term social adjustment, including social isolation and victimization.^{1, 2} Recent studies examining specific components of social adjustment, such as peer relationships, have found that survivors are less likely to name a friend than children with non-central nervous system tumors³ and are at high risk of having no reciprocated best friendship nominations in sociometric analyses.⁴ Caregiver ratings also highlight social concerns among survivors, with poor social outcomes rated as among the most impactful late effects in survivorship.⁵ This has prompted recent intervention efforts to support survivor social functioning.^6,7 However, much remains unknown with respect to predictors of social adjustment among survivors. Indeed, a recent review highlighted the identification of determinants of social impairment as one of the most important emerging areas in survivorship research.⁸

Given known cognitive deficits following cancer and cancer-related treatments,⁹ as well as the known interrelationship of cognitive and social functioning,^10,11 it is likely that cognitive skills are drivers of survivor social outcomes. Schulte and colleagues¹² found that cognitive impairment mediated the association between cranial radiotherapy and survivor social outcomes, including quality of social interactions and social withdrawal. Other researchers have examined more specific cognitive outcomes. For example, Dejardins and colleagues¹³ found an association between metacognitive skills and parent-rated survivor social skills (e.g., cooperation, assertiveness). Studies have also highlighted attention problems as significant predictors of survivor social challenges.^14,15 Continued investigation into known cognitive skill deficits among survivors (e.g., processing speed, executive functioning) and the impact of these deficits on aspects of social functioning is warranted, particularly given known associations between cognitive predictors and social outcomes in other pediatric traumatic and acquired brain injury populations^16–18 as well as autism spectrum disorder. ^{19, 20}

In addition to examining specific cognitive predictors, an investigation of survivor social outcomes should consider different radiation treatment modalities and the neurocognitive sparing potential of proton radiotherapy (PRT). Given the unique physical properties of PRT, maximum radiation dose is delivered to the tumor, with less entrance dose and no exit dose compared to photon radiation (XRT). ^21–23 Thus, with the sparing of more healthy brain tissue, PRT may yield better neurocognitive outcomes compared to XRT. The few comparison studies published to date suggest neurocognitive sparing with PRT, with evidence for higher IQ, processing speed, perceptual reasoning, executive functioning, and fine motor coordination among survivors treated with PRT versus XRT.^24–27 However, continued comparison of clinically meaningful outcomes across radiation groups is warranted.

Of these clinically meaningful outcomes, social functioning remains particularly understudied. If indeed PRT offers a neuroprotective benefit over XRT, it is plausible that better neurocognitive outcomes among survivors treated with PRT would support better social outcomes compared to those treated with XRT. However, this relationship between therapy modality, neurocognitive outcomes, and social functioning has not been directly examined. In their recent study comparing PRT and XRT treatment groups, Gross and colleagues²⁶ found better parent-rated social functioning for patients treated with PRT relative to XRT on a broadband adaptive skills rating measure. In another study examining quality of life more broadly, PRT patients were found to have higher quality of life ratings in the psychosocial domain relative to XRT patients.²⁸ While these findings are promising, cognitive predictors were not examined.

The present study offers the first known examination of a wide range of neurocognitive skills as predictors of peer relationships and social skills between pediatric brain tumor survivors treated with PRT versus XRT in late survivorship. We hypothesized that survivors treated with PRT would outperform those treated with XRT on cognitive measures and would be rated by caregivers as having better social outcomes. We also hypothesized that cognitive outcomes would predict survivor peer relations and social skills.

Methods

Participants

Participants were part of a larger study examining long-term neurocognitive outcomes in pediatric brain tumor survivors. Participants were enrolled according to the following eligibility criteria: (1) treated with a single course of PRT or XRT for a primary brain tumor, (2) no evidence of active disease at enrollment, (3) age ≥ 6 years at evaluation, and (4) fluent in English. No participants were receiving treatment for recurrence at the time of participation. All participants treated with XRT were treated between 2001 and 2006, while those treated with PRT were treated between 2007 and 2013. The timing defined for the two RT groups differed due to a shift from XRT to PRT as standard of care in 2007 at our institution. Eligible participants for the parent study were identified via medical record review and were approached for enrollment between 2011 and 2018. An 87% participation rate was achieved. Patients who declined participation did not significantly differ from enrolled participants based on RT type, sex, race, or histology (data not shown, all p>0.05). Of note, given our interest in long-term neurocognitive and social outcomes, patients diagnosed with brain stem glioma, high-grade glioma, or atypical teratoid/rhabdoid tumors were excluded from participation. Data were also excluded for participants who could not complete testing due to profound cognitive or visual impairment (n=5). The present study reports on the outcomes of 58 patients. Sample characteristics are reported in Table 1.

TABLE 1.

Demographic and clinical characteristics by radiation type

	XRT (n = 20)		PRT (n = 38)
	n	%	n	%	p
Sex					0.35
Male	16	80	26	68
Female	4	20	12	32
Race/Ethnicity					0.44
White	11	55	21	55
Black	2	10	4	11
Hispanic/Latino	7	35	9	24
Other	0	0	4	10
Maternal Ed.					0.43
< High School	2	10	3	8
High School	11	55	13	34
College Degree	6	30	16	42
MA or PhD	1	5	5	13
Unknown	0	0	1	3
Insurance Status					0.70
Private	14	70	24	65
Publica	6	30	13	35
Family Income ($)					0.70
< 40,000	7	35	13	35
40,000-79,999	5	25	6	16
80,000 and above	8	40	18	49
Tumor Location					0.36
Infratentorial	12	60	18	47
Supratentorial	8	40	20	53
Tumor Type					0.25
Gliomas	3	15	9	24
Medulloblastoma/PNETs	10	50	15	40
Ependymomas	4	20	2	5
Germ Cell	1	5	7	18
Other	2	10	5	13
Craniotomy					0.65
Yes	19	95	33	87
No	1	5	5	13
RT Technique					0.73
CSI	12	60	21	55
Focal	8	40	17	45
Shunt					0.15
Yes	11	55	28	74
No	9	45	10	26
Chemotherapy
Yes	15	75	22	58	0.20
No	5	25	16	42
	Mean (sd)	Min-Max	Mean (sd)	Min-Max	p
Age at Diagnosis (years)	5.0 (2.2)	0.9 – 10.1	7.6 (3.9)	1.1 – 15.5	<0.01
Time Since RT (years)	8.7 (2.8)	4.0 – 13.9	4.6 (2.5)	1.2 – 8.7	<0.01
Age at Evaluation (years)b	14.3 (3.1)	8.5 – 19.0	13.0 (2.8)	8.5 – 19.0	0.10
Total RT Dose to Tumor (Gy)	53.0 (6.4)	30.6 – 59.4	52.1 (3.9)	45.0 – 59.4	0.51
Maximum Tumor Diameter (cm)	4.9 (1.4)	2.0 – 8.0	4.5 (1.7)	1.9 – 8.9	0.41
Percent Below Poverty Level by Zip Codec	12.8 (0.1)	N/A	12.6 (0.1)	N/A	0.94

Note. Abbreviations: Ed = education, PFS = posterior fossa syndrome, RT = radiation therapy, CSI = craniospinal irradiation; Demographic and clinical factors were compared by radiation type using Chi Square, Fisher exact test, or independent t-tests, as appropriate.

^(a)Public insurance included Medicaid as well as the Children’s Health Insurance Program (CHIP), both of which provide health insurance for children of families with low income.

^(b)The maximum age at evaluation was 18.95 for the XRT group and 18.98 for the PRT group. The age 19.0 years is displayed due to rounding.

^(c)This variable reflects the percent of households in poverty within the home zip code of a patient.

Measures

All participants completed a neurocognitive battery with standardized measures, and caregivers completed norm-referenced parent rating scales. Trained research assistants administered all measures in accordance with standardized procedures. Relevant variables for the present study are delineated in Table 2. The age-appropriate version was administered for those measures with different forms/versions based on age. Of note, the WISC-V, WISC-IV, and WAIS-IV correlate strongly on the indices included in the present analyses (r=0.77-0.91).^{29, 30} Participant peer relations and social skills were assessed using the Conners 3 and BASC-3 rating forms, respectively. The Peer Relations scale provides an estimate of the quality of friendships maintained by the child and the extent to which the child appears to be accepted by peers.³¹ The BASC-3 Social Skills scale provides an estimate of a child’s ability to interact successfully with peers and adults in the home, school, and community.³² Either the child or adolescent versions of the BASC-3 Parent Rating Scale were administered depending on the participant’s age; however, the scales and composites contain similar content at all age levels.³³ The following cognitive variables were assessed: processing speed, executive functioning, verbal learning, verbal memory, sustained attention, and overall intellectual functioning. Several standardized measures were administered to assess different aspects of executive functioning, including shifting and inhibition. Similar to previous studies,^{34, 35} these measures were combined into an executive function composite (see Table 2). The use of a composite variable was supported by research demonstrating a common factor underlying executive functions and variable differentiability for tasks measuring shifting and inhibition in children.^36–38 Standardized scores were computed using age norms for all measures. All evaluations were conducted prior to the onset of the COVID-19 pandemic; therefore, standardized administration procedures and childhood social opportunities were not impacted by changes in procedures and enhanced safety precautions resulting from the pandemic.

TABLE 2.

Measure descriptions

Domain	Measure(s)	Variable	Measure Description	Scale
Peer Relationships	Conners 3, Parent and Teacher Forms	Peer Relations Scale	Scale measuring perceived quality of child’s friendships, acceptance by peers, and overall social connectedness	T-Score M = 50, SD = 10
Social Skills	BASC-3 Parent Form	Social Skills Scale	Scale measuring perceived quality of child’s social skills and assessing for social difficulties	T-Score M = 50, SD = 10
Full Scale IQ	WISC-IV; WISC-V; WAIS-IV	Full Scale IQ	Composite measure of global intellectual ability	Standard Score M = 100, SD = 15
Processing Speed	WISC-IV; WISC-V; WAIS-IV	Processing Speed Index (PSI)	Composite measure of ability to complete simple graphomotor tasks quickly	Standard Score M = 100, SD = 15
Executive Functioning				Standard Score M = 100, SD = 15
	D-KEFS Trail Making Test	Number-Letter Switching	Measure of cognitive flexibility and ability to maintain cognitive sets within working memory
	D-KEFS Color-Word Interference	Inhibition	Task measuring inhibitory control
	D-KEFS Color-Word Interference	Inhibition/Switching	Task assessing cognitive flexibility and inhibitory control
Verbal Learning	CVLT-C; CVLT-II	Total Recall Trials 1-5	Ability to learn verbal information over multiple learning trials	T-Score M = 50, SD = 10
Verbal Memory	CVLT-C; CVLT-II	Long Delay Free Recall	Delayed recall of verbal information	Z-Score M = 0, SD = 1
Sustained Attention	CPT-II	Omissions	Measure of inattention	T-Score M = 50, SD = 10

Note. BASC-3 = Behavior Assessment System for Children, Third Edition; WISC-IV = Wechsler Intelligence Scale for Children, 4th edition; WISC-V = Wechsler Intelligence Scale for Children, 5th Edition; WAIS-IV = Wechsler Adult Intelligence Scale, 4th Edition; D-KEFS = Delis-Kaplan Executive Function System; CVLT-C = California Verbal Learning Test, Children’s Edition; CVLT-II = California Verbal Learning Test, 2nd Edition; CPT-II = Conners’ Continuous Performance Test, 2nd Edition.

Participants under the age of 17 received the WISC-IV or WISC-V and the CVLT-C. Participants 17 and older were administered the WAIS-IV and CVLT-II. Three measures from the D-KEFS were incorporated into the Executive Functions variable. Each subtest scaled score was transformed into a Standard Score to facilitate interpretation. The composite represents an average of the three D-KEFS scores.

Statistical Analyses

Demographic and clinical factors were compared by radiation type (XRT vs PRT) using Chi Square, Fisher’s exact test, or independent t-tests, as appropriate. Group differences in social and cognitive functioning were assessed using analysis of covariance (ANCOVA), covarying for the effects of time since radiation and age at diagnosis. Because peer relations and social skills did not differ by radiation type, remaining analyses examined the sample as a whole. One-sample t-tests compared peer relations and social skills with the normative mean of 50. Frequency data on the number of participants rated as having impaired social outcomes were examined, with impairment defined as scores falling 1.5 SD above the mean for peer relations (i.e. ≥ standard score of 65) and 1.5 SD below the mean for social skills (i.e. ≤ standard score of 35). Further, χ² analyses determined whether the percentage of participants with impairment exceeded expectation assuming normal distribution of scores (6.68%). Bivariate correlations examined relationships between social outcomes and demographic, clinical, and cognitive variables. Predictors were included in multiple linear regression analyses based on significant correlations with peer relations and social skills. Demographic and clinical variables that significantly correlated (p<0.05) with the specific outcome were included. Cognitive skills that significantly correlated with either peer relations or social skills were also included. Power analyses were conducted for each regression model.

Results

Demographic and Clinical Characteristics

Comparisons of demographic and clinical characteristics between survivors treated with XRT and PRT are displayed in Table 1. The XRT group was younger on average at diagnosis than the PRT group (p<0.01). Reflecting the study design, the XRT group was further out from RT at the time of neurocognitive evaluation relative to the PRT group (p<0.001). Both groups included predominantly male survivors, and the majority of primary caregivers completed high school at a minimum. The majority of survivors in both groups underwent craniotomy and similar rates were treated with craniospinal radiation.

Cognitive and Social Outcomes in Survivors Treated with XRT versus PRT

Comparisons of social and cognitive outcomes between survivors treated with XRT and PRT accounting for time since radiation and age at diagnosis are presented in Table 3. Peer relations and social skills were similar across RT groups (all Fs<1, ps>.8). The average IQ score of the XRT group fell in the Below Average range and was significantly lower than that of the PRT group, which was Average (p<0.01). Cognitively, the XRT group performed significantly worse than the PRT group on measures of processing speed (p=0.01) and verbal memory (p< 0.01). Effects for executive functioning (p=0.07) and verbal learning (p=0.08) approached significance. Survivors treated with XRT and PRT performed similarly on a measure of sustained attention.

TABLE 3.

Social and cognitive skills by radiation type

	XRT (n = 20)			PRT (n = 38)
	Madj (SE)	Mean (SD)	Min-Max	Madj (SE)	Mean (SD)	Min-Max	F	p	η2
Social
Peer Relations	58.41 (4.91)	62.60 (19.77)	41 - 115	57.05 (3.31)	54.84 (18.10)	41– 104	.04	.84	<.01
Social Skills	51.04 (2.86)	50.74 (8.64)	33 – 68	50.88 (1.87)	51.03 (11.55)	28 - 68	<.01	.97	<.01
Cognitive
Full Scale IQ	75.68 (4.46)	74.20 (16.97)	44 – 101	94.33 (3.01)	95.11 (16.55)	59 – 134	9.98	<.01	.16
Processing Speed	71.56 (3.28)	68.85 (12.88)	50 – 91	82.52 (2.21)	83.95 (12.54)	65 – 112	6.40	.01	.11
Executive Functioning	81.76 (4.07)	77.92 (15.98)	55 – 101.7	91.66 (2.74)	93.68 (15.52)	56.7 – 120	3.38	.07	.06
Verbal Learning	85.54 (4.99)	84.45 (20.88)	44 – 116	97.44 (3.43)	98.03 (18.77)	55 – 133	3.20	.08	.06
Verbal Memory	84.12 (4.31)	82.50 (22.09)	32 – 115	100.49 (2.91)	101.34 (13.97)	70 – 130	8.21	<.01	.13
Sustained Attention	65.50 (5.56)	66.67 (29.80)	42 – 150.1	55.64 (3.75)	55.02 (14.19)	40.1 – 97.8	1.79	.19	.03

Note. ANCOVAs included time since radiation therapy and age at diagnosis as covariates. Adjusted means adjust for each of these covariates. T-scores are presented for social variables and sustained attention; Standard Scores (M = 100; SD = 15) are presented for all remaining variables. For Peer Relations, higher scores indicate worse functioning. For Social Skills, lower scores indicate worse functioning. For the cognitive measures, lower scores indicate worse performance with the exception of sustained attention, where higher scores indicate cause for concern.

Social Skills and Peer Relations among Survivors

Because peer relations and social skills did not differ by radiation type, remaining analyses examined the sample as a whole. Survivors were rated as having more difficulties with peer relations (M=57.52, SD=18.89) relative to the normative mean of 50, t(57)=3.03, p<0.01, though increased difficulties were not observed for social skills (M=50.93, SD=10.59), t(56)=0.66, p=0.51. However, the percentage of participants meeting criteria for impairment in peer relations (28%) and social skills (12%) exceeded expectation (i.e., 6.68%), (χ²(1)=38.67, p<0.001; χ²(1)=5.63, p<0.05).

Item level analysis of the Peer Relations scale indicated that 22% of pediatric brain tumor survivors were “often” or “very often” not invited to play or go out with others, and 27% were among the last to be selected for teams or games. Further, parent ratings indicated that 9% of survivors had no friends and did not know how to make friends. Among those participants rated as impaired on the Peer Relations scale, 81% were rated as not being invited to play and 75% were picked last for teams/games. 25% were rated as having no friends and 31% were rated as not knowing how to make friends. On the Social Skills scale, approximately half of survivors overall (45-51%) were rated as “never” or only “sometimes” demonstrating prosocial social skills such as complimenting others, showing interest in the ideas of others, and volunteering to help. Among survivors rated as impaired, 71-100% were rated as “never” or only “sometimes” demonstrating these prosocial skills (see Table 4).

TABLE 4.

Parent-rated social challenges across individual items

	Whole Sample	Participants Impaired on Social Outcome Measure
Conners 3 Peer Relations Items – Parent Form	“Often” and “Very Often” # (%)	“Often” and “Very Often” # (%)
Does not get invited to play or go out with others	13 (22)	13 (81)
Has no friends	5 (9)	4 (25)
Has trouble keeping friends	2 (4)	2 (13)
Is one of the last to be picked for teams or games	16 (27)	12 (75)
Interacts well with other children	51 (88)	10 (63)
Does not know how to make friends	5 (9)	5 (31)
BASC-3 Social Skills Items– Parent Form	“Never” and “Sometimes” # (%)	“Never” and “Sometimes” # (%)
Compliments others	30 (51)	7 (100)
Encourages others to do their best	22 (38)	7 (100)
Congratulates others when good things happen to them	20 (34)	7 (100)
Says, ‘please’ and ‘thank you.’	12 (21)	5 (71)
Tries to bring out the best in other people	24 (41)	7 (100)
Shows interest in others’ ideas	26 (45)	7 (100)
Volunteers to help with things	27 (47)	7 (100)
Offers help to other adolescents/children	22 (38)	7 (100)

Note. BASC-3 = Behavior Assessment System for Children, Third Edition.The age-appropriate forms of the BASC-3 and Conners 3 were provided to parents. The column labeled “participants impaired on social outcome measure” includes only those participants whose overall score on the Conners 3 Peer Relations scale fell 1.5 SD above the mean (i.e. ≥ standard score of 65; n = 16) or whose overall score on the BASC-3 Social Skills scale was 1.5 SD below the mean (i.e. ≤ standard score of 35; n = 7). BASC-3 rating options include Never, Sometimes, Often, and Almost Always. Conners 3 rating options include Never/Seldom, Occasionally, Often/Quite a bit, and Very often/Very frequently.

Correlates of Social Functioning

Increased difficulty on the Peer Relations subscale was significantly correlated with greater percent below poverty level (p<0.01), longer time since radiation (p<0.05), and shunt placement (p<0.01; see Table 5). Increased social skill problems were significantly correlated with larger tumor diameter and public insurance (both p<0.05). Slower processing speed, reduced executive functioning, and poorer verbal memory correlated with increased peer relation difficulties. Reduced social skills significantly correlated with slower processing speed alone.

TABLE 5.

Clinical and cognitive variables and correlations with social outcomes

	Peer Relations	Social Skills
Sex	.05	.26
Race/Ethnicity	−.04	−.08
Maternal Education	.21	.24
Household Income	−.001	.107
Insurance	−.01	.28*
%Below Poverty	.41**	−.24
Tumor Location	−.03	−.20
Tumor Type	−.05	.16
Radiation Type	−.20	.01
Craniotomy	.13	.04
CSI	−.03	.00
Shunt	.34**	−.04
Chemotherapy	−.06	.17
Age at Diagnosis	−.14	.22
Time Since RT	.28*	−.06
Age at Evaluation	.23	.11
Total Tumor RT Dose	−.15	−.09
Maximum Tumor Diameter	.13	−.29*
Full Scale IQ	−.29*	.30*
Processing Speed	−.36**	.30*
Executive Functioning	−.43**	.26
Verbal Learning	−.23	.09
Verbal Memory	−.42**	.01
Sustained Attention	−.003	−.07

Note.

^*= p < .05;

^**= p < .01

Multiple linear regression models examined the independent and shared contributions of predictors that were significant in univariate correlations with outcomes (see Table 6). The overall peer relations model was significant, F(6,44)=3.76, p<0.01. Percent below poverty was a significant predictor, t=2.18, p<0.05, and verbal memory approached significance as a unique predictor, t=−1.99, p=0.05. Effect sizes for percent below poverty, f²=0.08, and verbal memory, f²=0.06, were modest. For social skills, the overall model was significant F(5, 50)=3.00, p=0.02, and tumor diameter emerged as the only significant predictor, t=−2.07, p<0.05. For the peer relations model, there was 80% power to detect a difference in R² of 0.107 for a predictor variable given the full model R²=0.34 (two-sided a=0.05). For the social skills model, there was approximately 80% power to detect a difference in R² of 0.113 for a predictor variable given the full model R²=0.23 (two-sided a=0.05).

TABLE 6.

Multiple linear regression models for social outcomes

Social Outcome	Effect	B	SE	Beta	p value	f2
Peer Relations (n = 51) R2 = 0.34 Adjusted R2 = 0.25	Time Since RT	.32	.80	.05	.70	.00
	% Below Poverty	46.53	21.30	.28	.03	.08
	Shunt	5.30	4.97	.14	.29	.02
	Processing Speed	.04	.23	.03	.86	.00
	Executive Functioning	−.66	1.01	−.13	.52	.01
	Verbal Memory	−4.13	2.08	−.28	.05	.06
Social Skills (n = 56) R2 = 0.23 Adjusted R2 = 0.15	Insurance	4.84	2.92	.22	.10	.04
	Maximum Tumor Diameter	−1.77	.86	−.26	.04	.07
	Processing Speed	.06	.15	.09	.67	.00
	Executive Functioning	.82	.63	.27	.20	.03
	Verbal Memory	−1.17	1.23	−.14	.35	.01

Discussion

Consistent with previous studies, the present findings highlight social difficulties among pediatric brain tumor survivors. Peer relationships (i.e., friendship quality, peer acceptance) emerged as an area of particular concern. Significantly more survivors were rated as having peer relationship challenges relative to normative expectation, with a sizable portion (~25%) experiencing social exclusion (e.g., not being invited to play, selected last for teams/games). Among those with impaired parent-reported peer relations, 81% were rated as not being invited to play with others. These findings broadly correspond with previous research highlighting survivor vulnerability to peer exclusion and isolation.^{39, 40} Further, a relatively small but remarkable number of survivors overall (~10%) were rated by caregivers as having no friendships. Within the impaired group, one-quarter were rated as having no friendships. These results are similar to previous findings,^{3, 41} although Hocking and colleagues reported a greater proportion of survivors (38%) who were unable to name a friend.

The proportion of survivors experiencing problems with social skills also significantly differed between survivors and normative expectations, which is consistent with previous findings documenting social skill deficits in this population.^{2, 40} Specific prosocial skills, such as complimenting others and offering to help, emerged as areas of difficulty on individual item review. With the exception of saying “please” and “thank you,” all survivors rated as having impaired social skills had difficulty with all prosocial skills evaluated. Ultimately, the results encourage continued efforts to better understand and support survivor social functioning.

The current findings identify several cognitive variables as potentially influential for social outcomes among survivors. Processing speed, an area in which survivors consistently demonstrate impairment, ^42–44 significantly correlated with both peer relationships and social skills. Executive functioning and verbal memory also correlated significantly with peer relations. However, no cognitive variables significantly predicted social skills, and only verbal memory approached significance as a unique predictor of peer relations in the context of other cognitive, treatment, and sociodemographic variables. The fact that verbal memory approached significance as a unique predictor of peer relations is not surprising given the role of verbal memory in social communication, facilitating the retrieval of previously learned information and incorporation of such information with conversational demands.⁴⁵ However, while verbal memory correlates with social problem solving and social skills in other medical populations (e.g., traumatic brain injury⁴⁶), the relationship between verbal memory and social outcomes has not been closely examined in the pediatric brain tumor population. As verbal memory is known to be at risk in survivors, likely as a result of disease and treatment-related damage to temporal structures and cortical connectivity underlying memory function,⁴⁷ this relationship warrants further investigation.

It is notable that survivor social outcomes did not differ based on RT modality, although significant differences in cognitive outcomes between survivors treated with XRT versus PRT extend previous studies suggesting a potential neuroprotective benefit of PRT for cognitive skills.^24–27 Given previous associations of cognitive impairment with poor social outcomes,¹² the XRT group was expected to have significantly lower social outcomes than the PRT group as a function of lower cognitive performance. Indeed, Gross and colleagues²⁶ identified significant differences between radiation groups on a measure of adaptive social functioning, suggesting that survivors treated with PRT have more favorable social outcomes relative to those treated with XRT. However, the present findings suggest that RT modality may be a less important driver of long-term social outcomes than other treatment and sociodemographic variables. Longer time since RT and shunt placement were significant correlates of survivor peer relation difficulties, and tumor diameter emerged as a significant predictor of social skills. These results echo early findings of worse parent-rated social skills for survivors with greater time since diagnosis⁴⁸ and suggest that disease factors and illness complications (e.g., tumor size, hydrocephalus) impact long-term social outcomes, potentially more so than RT modality. Furthermore, a higher percentage of household poverty predicted worse peer relationships. This finding reflects previous work identifying socioeconomic risk factors (e.g., neighborhood poverty, perceptions of neighborhood safety) as predictive of adverse peer relationships and other socio-emotional challenges.^49–51 Given these findings as well as the fact that both regression models, while significant, accounted for only a modest amount of variance, future investigations will benefit from considering specific treatment factors in addition to broader aspects of brain tumor diagnosis/treatment (e.g., school absences, internalizing and externalizing symptoms, family dynamics) and sociodemographic context as potentially meaningful influencers of survivor social outcomes.

The present study has several clinical implications. First, the social challenges observed among survivors were documented at long-term follow-up, indicating that late effects of brain tumor diagnosis and treatment likely include social deficits that occur well into survivorship. Further, the finding of social difficulties among survivors in both RT groups supports the need for careful monitoring of social adjustment regardless of radiation modality. Models and guidelines for the neuropsychological care of survivors⁵² should therefore emphasize continued surveillance and monitoring of social functioning, even in light of potential neurocognitive sparing associated with PRT. Additionally, results suggest that survivors with cognitive impairments at follow-up, particularly in verbal memory, may be at heightened risk for social difficulties. Survivors facing hardships associated with socio-economic disadvantage may also be at greater risk for poor peer relationships. It is also notable that peer relations, not social skills, emerged as the greater area of concern for survivors. This encourages careful consideration of assessment methods, with measures that directly inquire about friendships and social acceptance potentially capturing a broader picture of survivor social functioning than those evaluating social skills. Direct questions regarding friendships in an interview format, such as the approach utilized by Hocking and colleagues, ³ may also provide opportunities to offer clarification and thereby identify a greater number of survivors with social difficulties compared to parent ratings alone.

Limitations

Readers should note several study limitations. Given the practical and ethical barriers preventing a randomized controlled trial in this sample, patients were not randomized to RT groups. Additionally, RT groups differed with respect to the follow-up interval, as PRT patients were treated more recently than XRT patients. To minimize this difference, we examined the last available cohort of XRT patients and the first available cohort of PRT patients; however, there is still the potential that unmeasured treatment differences across groups (e.g., reductions in boost volume, advancements in surgical procedures) as well as potential differences in survivor support services influenced outcomes. The cross-sectional study design is another study limitation, as this precludes direct analysis of changes in cognitive and social functioning over time. Additionally, pre-RT neuropsychological assessment and comprehensive data regarding educational services/interventions were not available for this cohort. It should be noted that families with greater concern regarding their child’s outcomes and families who are not connected with supportive services might be more likely to remain engaged in follow-up through pediatric oncology centers, potentially resulting in a sample with more cognitive or functional difficulties. The small sample size is another limitation, likely affecting power to detect significant differences and predictors. The sample was also heterogeneous for tumor type. Although tumor type and location did not significantly differ across radiation groups, future studies will benefit from the evaluation of social outcomes within disease-specific groups. Regarding measurement, the present study incorporated only parent report measures for the assessment of social outcomes. Although parent ratings provide meaningful information and are commonly used in clinical settings, other studies have noted the benefits of sociometric approaches, computer-based measures of social information processing, and interviews for the evaluation of survivor social outcomes.^{3, 4, 53, 54} The lack of self-report measures is another limitation and may lead to under- or overestimation of social challenges depending on caregiver awareness of social difficulties experienced by survivors outside of the home. Future studies will benefit from carefully selected methods for measuring survivor self-reported social functioning, including an interview format permitting clarification. ³

Conclusions

Regardless of radiation modality, pediatric brain tumor survivors are at risk for long-term social difficulties, with perceived friendship quality and peer acceptance being areas of particular vulnerability. Survivors are also at risk for cognitive late effects, and those with weaknesses in verbal memory may be at higher risk for peer relationship challenges. Treatment variables such as time since radiation, hydrocephalus/shunt placement, and tumor diameter may prove more influential for long-term social functioning than radiation modality, although further investigation is needed. Overall, survivors will benefit from continued monitoring of cognitive and social functioning over the course of survivorship. It is hoped that further inquiry into cognitive and clinical predictors of social outcomes will inform interventions to support survivor social adjustment and overall quality of life.

Abbreviations

PRT

Proton Radiation Therapy

XRT

Photon Radiation Therapy

RT

Radiation Therapy

CSI

Craniospinal Irradiation

IQ

Intelligence Quotient

ANCOVA

Analysis of Covariance

BASC

Behavior Assessment System for Children

WISC

Wechsler Intelligence Scale for Children

WAIS

Wechsler Adult Intelligence Scale

D-KEFS

Delis-Kaplan Executive Function System

CVLT-C

California Verbal Learning Test, Children’s Edition

CVLT-II

California Verbal Learning Test, 2nd Edition

CPT-II

Conners’ Continuous Performance Test, 2nd Edition

Data Availability Statement

The data that support the findings of this study are available from the corresponding author upon reasonable request.