The purpose of the current study was to investigate longitudinally children's behavioral and social competence outcomes up to 2 years after pediatric stem cell transplantation (SCT) and related factors.
The purpose of the current study was to investigate longitudinally children's behavioral and social competence outcomes up to 2 years after pediatric stem cell transplantation (SCT) and related factors.
Ninety-nine mothers and 24 youths completed standardized questionnaires (Child Behavior Checklist [CBCL] and Youth Self-Report [YSR]) pre-SCT, and 12 and 24 months after SCT; 26 teachers completed Teacher Report Form (TRF) at 24 months. Information regarding clinical (eg, diagnosis, cranial radiation, graft vs host disease [GVHD]), child (eg, age, sex, physical health), and familial (eg, maternal age, education, distress) factors was also obtained.
Linear mixed regression models with compound covariance structure followed by adjusted pairwise analyses yielded significant improvements from pre-SCT to 1 and 2 years after SCT in total scores; in externalizing and internalizing scores from pre-SCT to 2 years after SCT; and in total competence from pre-SCT to 1 year after SCT. Child's physical health, maternal age, and depression were found to be significantly associated with the total, internalizing, and externalizing CBCL scores, whereas GVHD, mother's age, and time since diagnosis were associated with CBCL competence scores. Diagnosis, cranial radiation, GVHD, child's physical health, and maternal age and education were associated with YSR total behavioral and competence scores. Finally, total TRF scores were associated with time since diagnosis; TRF educational and adaptability scores were associated with maternal education, age, and distress.
Clinical, personal, and familial factors must be considered to understand the psychosocial outcomes of these survivors up to 2 years after SCT. This study has important implications for psychosocial interventions for this population. Cancer 2009. © 2009 American Cancer Society.
It is now well established that pediatric blood transplant protocols, including bone marrow, hematopoietic progenitor, or stem cell transplantation (SCT), are aggressive yet life-saving procedures for children diagnosed with some cancers and blood and metabolic disorders. Although medical efficacy continues to improve, there are still concerns with physical and psychosocial sequelae related to the original disease or complications associated with SCT1, 2 and increased neurocognitive deficits3 and mortality in long-term pediatric and adult survivors of SCT.4 Children who qualify for SCT may have already endured years of aggressive treatments for a life-threatening illness, disease recurrence, and related psychosocial difficulties. The stress associated with SCT may increase their risk for additional psychosocial maladjustment. However, the literature on the psychosocial effects of SCT is inconsistent. The main purpose of this study was to describe longitudinally psychosocial adjustment of survivors of pediatric SCT, pre-SCT and 1 year and 2 years post-SCT.
In a preliminary report, we found that compared with a normative sample, patients presented with no behavioral problems, either internalizing or externalizing, at pre- and 6 months post-SCT.5 However, with a larger sample, children who have been treated for acute lymphoblastic leukemia (ALL) had the poorest psychosocial scores compared with norms for healthy peers pre-SCT.5 Shortly after the procedure, children may become withdrawn and passive, with increased symptoms of anxiety and depression.4-7 Some have reported behavioral problems 6 months after SCT.8, 9 In general, however, survivors are reported to adjust well.6, 8, 10, 11 A small subgroup (15%-30%), nonetheless, are reported to experience late neurocognitive,12 psychologic, and/or social difficulties11, 13; increased social isolation14; and poor self-esteem and emotional well-being.7, 8
Parents and teachers tend to report more behavioral problems than survivors themselves.15, 16 Although in a study of 13 survivors who completed questionnaires between 6 months and 8 years after SCT, 58% reported difficulties returning to school, and 25% identified problems making new friends.17 In another study, the majority of adolescent survivors reported lack of self-confidence and energy, moodiness, feeling worried about their appearance (short stature/disfigurement), feeling that a significant part of their life has been missed, and anxiety regarding the possibility of recurrence of disease.18
The relation between psychosocial adjustment of SCT survivors and disease-treatment variables, child characteristics, and familial factors is not well documented. We and others found no association between severity of symptoms pre-SCT and behavior at 6 months19 and 12 months after SCT.15 However, greater internalizing and externalizing behavioral problems and lower school competence compared with normative samples were reported years after transplant by survivors who endured cranial radiation.16, 20, 21
In terms of the child's characteristics, younger children are reported to have greater difficulty returning to school and engaging in social activities than older children,11 but neither age nor sex differentiated psychosocial outcomes in other studies.11, 13, 16, 19 With regard to familial factors, lower socioeconomic status has been associated with lower social competence, and family expressiveness and cohesion have been associated with fewer behavioral problems and better self-competence 6 months after SCT.8, 19
In summary, the literature is inconclusive regarding the short-term and long-term psychosocial outcomes of pediatric SCT and factors that might be related to these outcomes. Differences in findings may be because of methodologic inconsistencies and limitations such as small sample size, different types of informant (child, parent, other) or data-gathering tools (standardized questionnaires, open-ended interviews), and wide variations in the time at which survivors were tested (ie, 6 months to >10 years after transplant). To our knowledge, very few longitudinal studies have been conducted to date, masking the differential effects of changes related to the transplant procedure over time and the age effect on the outcomes. Finally, rarely have previous studies used a conceptual model to guide their empirical findings.
We have adapted Wallander and Varni's22 multivariable model for understanding the impact of childhood chronic illnesses on the child's psychologic adjustment to investigate the role of child characteristics and familial factors over and beyond the effect of childhood cancer with children23 and parents.24 We also examined the aggregated effect of previous disease and SCT variables, and child and familial factors on the child's cognitive, educational, and visual-motor functioning.25 On the basis of this model, our previous findings,25 and current empirical literature, in the present longitudinal study we considered a child's previous disease and SCT experience as potential risk factors that could have a detrimental effect on the child's psychosocial adjustment, alone or in combination with child characteristics and family factors. Both child characteristics and family factors may serve as resistance factors on the child's psychosocial adjustment. Specifically, using several informants, we investigated psychosocial adjustment of survivors over time and its correlates (clinical factors: diagnosis before SCT, cranial radiation, graft vs host disease [GVHD], time since diagnosis; familial characteristics: maternal age, education, distress, family cohesion; and child's characteristics: age, sex, physical health).
This study is part of a large project on the psychosocial outcomes after pediatric SCT. The method has been described in detail elsewhere.25 Briefly, mothers of 99 patients (of 120 potential families) and 24 youths (of 29 who were old enough [aged ≥11 years] to complete the self-report measure) participated is this study. Scheduling conflicts or changes in the child's health status (disease recurrence) were reasons for not participating. At 1 and 2 years after SCT, 50 and 46 mothers and 16 and 17 youths, respectively, completed the measures. This attrition was because of mortality (40%), poor health (16%), difficulties scheduling or conducting assessments because of families returning to their original treating center for follow-up (38%), and loss of interest in the study (6%). This attrition rate is not uncommon in longitudinal studies with SCT survivors.26 Table 1 presents clinical, child, and familial characteristics of the sample. The majority of children were diagnosed with ALL or other leukemia types (ie, acute myelogenous leukemia). Their mean age pre-SCT was 8.27 years (standard deviation [SD], 4.38), and the mean time since diagnosis was 19.45 months (SD, 23.74). Most children came from a 2-parent Caucasian family and a middle-class socioeconomic background.27
|Age at transplant, y||8.27||4.38|
|Diagnosis prior to transplant|
|Other solid tumor||17||17.17|
|Type of transplant|
|Graft versus host disease|
|CHQ physical health|
|Up to high school||31||31|
|Some college or university||22||22|
|College or university completed||46||46|
|Maternal depression scores|
After approval from the research ethics board was obtained, potential participants were invited by mail to participate. Informed consent and assent were obtained at the initial assessment. Mothers of all participants and youths aged ≥11 years completed the Child Behavior Checklist (CBCL)28 and the Youth Self-Report (YSR),29 respectively, and teachers completed the Teacher Report Form (TRF)30 at the last assessment for children who were attending school. The CBCL for ages 4 to 18 years is a parent-completed questionnaire, used to assess psychosocial adjustment and social competence. It consists of 112 items, which assess common problem behaviors and social competence in terms of the child's everyday activities, friendships, and educational achievement. Ratings are summed to obtain total, internalizing (symptoms of anxiety and depression), and externalizing (symptoms of aggression) scores for the behavioral scale and total competence, activity, social, and educational scores for the competence scale. All scales have good test-retest reliability (r = 0.89, 0.93, and 0.93, respectively) and adequate validity. The CBCL has been used in previous studies of pediatric cancer patients who undergo SCT.11, 16, 19, 23, 31-33 Although the CBCL has been criticized for its use to assess behavioral problems and social competence in children with chronic illnesses,34 we used this measure because 1) it facilitates comparisons across studies; and 2) in real life, survivors and others use peers as reference for comparison. The YSR for ages 11 to 18 years consists of 118 items and, similar to the CBCL, assesses common problem behaviors and social competence. The summary test has good test-retest reliability assessed at 1 week (r = 0.82-0.92) and 6 months (r = 0.62-0.74) and adequate validity.29 Finally, the TRF includes 113 behavioral items similar to the CBCL and has satisfactory psychometric properties. TRFs were completed for 25 participants at 2 years post-SCT only because many children did not consistently attend school until after their first year after SCT. For each version, raw scores are converted to standardized T scores, with a uniform mean of 50 and SD of 10 for the behavioral scales; scores between 60 and 70 are considered borderline, and scores ≥70 are considered to be within clinical range. For the competence scales, T scores >33 are considered within the normal range, those between 33 and 30 within the borderline range, and scores <30 within the clinical range.
The Child Health Questionnaire,35 in its 50-item parent-report version, is a generic, well-validated measure designed for use across diverse groups of children with various health conditions. It provides 2 summary scores, physical and psychosocial health. We used the physical summary scores to assess the child's physical health. The Family Adaptability and Cohesion Evaluation Scale (FACES-III),36 measuring family cohesion, and the Beck Depression Inventory (BDI),37 assessing maternal distress, were used to investigate potential related factors. Both measures are brief and widely used with families affected by childhood chronic illness or cancer and could be completed by a parent with minimal burden. FACES-III is a 20-item scale with 2 dimensions of family relations: adaptation and cohesion. Test-retest reliability over a 4-week to 5-week period was r = 0.80 for adaptability, and r = 0.83 for cohesion.36 We used the cohesion score as a measure of family functioning. The BDI is a self-administered 21-item questionnaire designed to measure depressive symptomatology. The BDI has been found to be highly sensitive in measuring change and severity of depressive symptoms.37 Concurrent validity of the BDI with other self-report measures of depressive symptoms has been consistently high.37 Both measures were completed by the mother.
The child's disease and treatment variables were obtained from the medical chart. These included: diagnosis type (ALL, other leukemias, neuroblastoma, other tumors, and hematological disorders); time since diagnosis at pre-SCT (<6 months, between 6 months and 1 year, and >1 year); radiation treatment (no radiation, cranial radiotherapy [CRT] before SCT, total body radiation [TBR] as part of the conditioning regime for SCT, and combined CRT and TBR). GVHD was categorized into “yes” or “no,” based on reported presence or absence of disease. The child characteristics age at SCT and sex, and maternal age and education at SCT were also included, with the latter as an index of socioeconomic status.
SAS statistical software (version 9.1; SAS Institute Inc. Cary, NC) was used for the statistical analysis. Descriptive statistics were calculated for the outcome variables and covariates of interest. The Student t test for independent samples was used to compare pre-SCT data of participants who continued in the study with those who did not. To avoid collinearity, a pairwise assessment of associations between predictors was examined. Chi-square tests within categorical variables, analysis of variance for continuous and categorical variables, and correlation analyses for continuous variables were conducted. Variables found to have high potential collinearity were not included together in a multivariable model. Bivariate analyses were conducted for the outcomes and their covariates at each assessment time. Covariates significant at the 20% level in the bivariate analyses were included in the full multivariable linear mixed model analyses. The linear mixed model with a compound symmetry covariance structure is a multivariable regression model that tests for repeated measures within optimum use of all available data at each assessment time, and accounts for collinearity of the repeated measures. Nonsignificant (P > .05) variables were removed from the models through backward selection. For some analyses, BDI scores were stratified into <17 and >17 using the cutoff for clinical symptoms of depression. The linear mixed model analyses tested the longitudinal effect of time and the effect of clinical, child, and family factors on each of the behavioral outcomes based on maternal and self-reports. Pairwise analysis was conducted using the Tukey-Kramer adjustment for post hoc multiple comparisons. To examine relationships with behavioral scores based on teachers' reports 2 years after SCT, we planned to conduct generalized linear regression analyses. However, this was not possible, because preliminary analyses indicated that these data were not normally distributed. Therefore, we conducted nonparametric tests and Spearman correlations to assess associations with teachers' data. To determine clinical significance, effect size (ES) was calculated. An effect size of 0.50 corresponds to a minimal perceptible difference and hence could be considered of clinical significance.38
Comparisons of participants who completed the follow-up with those who only completed the pre-SCT assessment yielded no significant differences in any of the variables, except for radiation treatment. More children with a history of cranial radiation before transplant did not continue with the study (chi-square, 6.80; P < .05). Table 2 presents the means and standard deviations of the total, externalizing, and internalizing T scores for the CBCL, YSR, and TRF, as well as social and educational scores over time. All group means were within the normal range regardless of time of assessment or informant. However, at pre-SCT, 26%, 33%, and 19% of CBCL total, internalizing, and externalizing mean T scores, respectively, were >60, borderline level, or worse. By 2 years after SCT, the percentage of T scores >60 was reduced to 13% for each of the CBCL total, internalizing, and externalizing scores. The percentage of YSR T-scores at the borderline level at pre-SCT were not as high (20%, 12%, and 12% for total, internalizing, and externalizing, respectively), and only a small percentage of survivors had scores >60 at 2 years post-SCT for internalizing (12%), with none for externalizing or total scores.
|Outcomes||Pre-SCT||1 Year After SCT||2 Years After SCT|
There was a time effect on the CBCL total scores (F = 8.85; P = .0003), suggesting a reduction of behavioral problems over time as perceived by the mother. By using the Tukey-Kramer test to adjust for multiple comparisons, we found a significant reduction in the scores from pre-SCT to 1 year after SCT (t112 = −2.38; adjusted P = .05; ES = 0.22), and from pre-SCT to 2 years after SCT (t111 = −4.13; adjusted P = .0002; ES = 0.48), although the effect size was small to medium. No significant change was found between 1 and 2 years after SCT. Similarly, analyses of the CBCL externalizing (F2109 = 5.29; P = .0064) and internalizing scores (F2,112 = 8.40; P = .0004) also indicated significant time effects. Adjusted pairwise comparisons demonstrated significant reduction of externalizing scores from pre-SCT to 2 years after SCT, with a small effect size (t112 = 3.15; adjusted P = .0059; ES = 0.15), but not from pre-SCT to 1 year after SCT. The change from pre-SCT to 2 years after SCT was significant for the internalizing scores (t118 = −4.07; adjusted P = .0003; ES = 0.60), and the effect was medium to large, suggesting that mothers perceived an overall improvement in internalizing behavior by 2 years after SCT. CBCL total social competence scores also yielded a significant time effect (F2,53 = 4.19; P = .02). Pairwise comparisons demonstrated significant improvement from pre-SCT to 1 year after SCT, with a small effect size (t59 = −2.88; adjusted P = .02; ES = 0.15), but not from pre-SCT to 2 years after SCT. The YSR behavioral and competence scores yielded no significant time effect, although there was a reduction in the behavioral means of at least 10 points from pre-SCT to 1 year after SCT.
The mixed regression model found child's physical health (F1,90 = 3.80; P = .05) and maternal depression symptoms (F1133 = 5.92; P = .0163) to be significantly associated with total CBCL scores. Survivors with the worse physical health and whose mothers reported more depression symptoms had higher total CBCL scores compared with other survivors. Similar results were found by stratifying maternal depression symptoms into <17 and >17. With the CBCL externalizing scores, maternal age (F1,99 = 9.12; P = .0032) and maternal depression symptoms (F1193 = 14.47; P = .0002) were significantly and strongly associated, indicating that externalizing scores were higher when mothers were younger and had more symptoms of depression. With CBCL internalizing scores, only mother's depression symptoms (F1122 = 6.54; P = .0118) were significantly associated, suggesting that mothers who had more depression symptoms rated their children as having high internalizing scores. GVHD (F1,50 = 3.74; P = .05) and maternal age (F1,53 = 5.70; P = .02) were found to be significantly associated with the CBCL competence scores, suggesting that children's total competence scores were lower when mothers were younger and children developed GVHD. Moreover, mother's age was significantly associated with social scores (F1,59 = 5.04; P = .03), indicating that children with higher social scores had older mothers. Finally, child's age (F1,61 = 6.86; P = .01) and time since diagnosis (F1,59 = 11.83; P = .001) were both significantly associated with educational scores, suggesting that the younger the child at pre-SCT and the longer the time since diagnosis, the worse the educational scores as reported by the mother.
GVHD (F1,24 = 4.84; P = .038), diagnosis (F4,25 = 3.74; P = .027), mother's education (F2,27 = 6.30; P = .009), and mother's depression symptoms (F1,29 = 8.97; P = .006) were found to be significantly associated with the YSR total scores. Children who developed GVHD had higher YSR total scores than those who did not. The adjusted Tukey-Kramer test indicated that those with a diagnosis of ALL reported higher total YSR scores compared with those with solid tumors, with a large effect size (t17 = 3.61; P = .019; ES = 1.36). Moreover, survivors whose mothers had more depression symptoms also had higher total scores. Only radiation treatment was significantly associated with externalizing scores (F2,29 = 7.97; P = .0018), suggesting that survivors who received both CRT and TBR had higher externalizing behavior problems than those survivors who had only TBR (t33 = 3.99, adjusted P = .001; ES = 0.98) or had none (t38 = 2.33; adjusted P = .06; ES = 0.54). Similarly, cranial radiation treatment (F2,31 = 4.93; P = .0139) and physical health (F1,48 = 6.92; P = .0114) were found to be significantly associated with internalizing scores. Thus, survivors who had both CRT and TBR and worse physical health obtained higher internalizing mean scores than other survivors (t33 = 3.13; adjusted P = .01; ES = 0.79). These effects were medium to large. Moreover, maternal education was significantly associated with YSR total competence scores (F2, 27 = 3.81; P = .034), with survivors with higher competence scores having mothers who completed a university education.
Time since diagnosis at pre-SCT was found to be significantly correlated with total TRF and internalizing scores (Rho = 0.43; P = .03; Rho = 0.46; P = .03, respectively), indicating that the longer time since diagnosis, the higher the internalizing scores. Maternal age at pre-SCT was also associated with TRF academic, learning, and total adaptability scores (Rho = 0.45; P = .03; Rho = 0.44; P = .03; and Rho = 0.41; P = .04, respectively). In every case, the older the mother was the better the TRF scores. Finally, maternal symptoms of depression were associated with TRF learning scores (Rho = 0.41; P = .04), indicating that children whose mothers had more symptoms of depression were rated by their teachers as having more learning problems. In all cases the strength of correlations were considered moderate to large, based on guidelines to assess strength of significant correlations.39
This longitudinal study investigated psychosocial outcomes of children and youths pre-SCT and 1 and 2 years after SCT, and the clinical, child, and familial factors associated with these outcomes as reported by mothers and survivors. It also investigated outcomes and related factors as reported by teachers 2 years after SCT. Although the mean behavioral scores, as reported by mothers and youths, were within the normal range at pre-SCT and remained so at 1 and 2 years after SCT, significant improvements were noted by mothers over time. This was also reflected in the reduction of survivors' internalizing scores that fell within borderline and clinical levels (33% at pre-SCT; 13% 2 years after SCT). There was also a perceived improvement in survivors' social competence at 1 year after SCT, but this returned to the pre-SCT level by 2 years after SCT, suggesting that when survivors' daily life was fully normalized by 2 years after SCT, mothers may have realized that their children were not as socially engaged (sports, clubs) as their peers. Although the psychosocial outcomes based on youth reports did not appear to change significantly over time, there was a reduction in the scores in the same direction as those reported by mothers. The lack of significance may be because of low statistical power and the tendency of youths to report fewer problems than parents.16, 17 Thus, although normative behavioral scores were previously found at 6 months after SCT with the same population19 and reported by others 2 years after SCT,11 this study not only confirms those results but goes beyond by providing a greater understanding of subtle behavioral improvements and longitudinal changes, as reported by youths and mothers. This highlights the importance of considering both perspectives when assessing the psychosocial impact of pediatric SCT.
On the basis of the conceptual model, some similarities and some differences across informants emerged in the factors identified in relations with psychosocial adjustment outcomes. In terms of similarities, we found that both mothers and youths reported a positive relation between high maternal distress (measured by symptoms of depression) and child's internalizing difficulties. In addition, child's poor health was related to child's behavioral problems as reported by mothers and internalizing difficulties as reported by youths. Moreover, both maternal and youth reports of low total competence were associated with the presence of GVHD (and low maternal education in youth self-reports). Consistency across informants was also found with the relation between longer time since diagnosis and having educational problems based on mothers' and teachers' reports, confirming the adverse effect of illness chronicity previously observed on the survivors' educational performance.25 Moreover, supporting previous results,25 mothers reported better educational outcomes in older children, suggesting that being older at SCT may act as a resistance factor for educational outcomes. The consistency across informants in this study and across studies regarding the relationships with cognitive and educational outcomes validates the findings and identifies maternal distress, the child's younger age, poor health, and long time since diagnosis as potential risk factors for psychosocial and educational outcomes of SCT survivors.
In terms of differences, although maternal age (younger) and distress (higher) were found to be related to maternal reports of children's externalizing problems, having a history of combined CRT and TBR and poor physical health were found to be related to youth reports of externalizing problems. These youth reports are partly consistent with reports that have identified teens with more aggressive treatments experiencing more adjustment problems.16, 18 Thus, it appears that mothers rated their children's acting out behavior as a function of their own distress and maturity, whereas children may have attributed the same behaviors to their treatment and poor health. This conclusion is supported by additional relations between youth reports of behavior problems and other clinical variables such as the type of cancer (ALL) and the presence of GVHD.
To summarize, over time, mothers perceived improvement in survivors' behavioral and social competence. By using maternal reports, the conceptual model best explains the psychosocial outcomes of survivors of SCT by considering clinical factors (time since diagnosis and GVHD), child's characteristics (age and physical health), and family factors (maternal age and distress). On the basis of youth self-reports, the clinical factors contributed best to explain their psychosocial adjustment, although maternal education and distress were also identified to influence the outcome. Finally, based on teacher reports the family factors best explained educational and psychosocial adjustment of the child. These findings stress the importance of considering the perspectives of children, parents, and teachers when assessing the impact of the SCT experience on survivors' psychosocial functioning and educational achievement and for planning psychosocial interventions.
Several limitations were identified in this study. A small sample size, particularly for the youth and teacher reports, lowered the statistical power for hypothesis testing, increasing the chances of type 1 error. There may be some sample selection bias based on survivors' health. However, as the results show, that does not appear to be the case. Mothers completed both the CBCL concerning their children and the depression inventory about themselves, which potentially could have biased the perceptions and rating of their children. The finding that both the youth and maternal ratings of behavior were associated with maternal distress, however, adds validity to maternal reports. The use of the CBCL to measure psychosocial adjustment and social competence with populations affected by a chronic illness requires caution,34 particularly when assessing social competence. Given the small number of items in this area, it is possible that the results reflect limited access to social situations because of frequent absences from school or physical impairments rather than actual deficits in social competence. Conversely, we can argue that despite this limitation, the instrument detected some differences and relationships that may suggest a conservative estimate. In addition, the wide use of this measure not only allows us to compare across studies, but serves as a benchmark of the expectations placed on survivors in their daily life. Finally, another limitation of this study is its relatively short follow-up period (2 years) when compared with recent follow-ups, suggesting that some subgroups of pediatric SCT survivors continue to have neurocognitive deficits and a higher mortality rate.3, 12
Despite these limitations, this longitudinal study provides new insight on the psychosocial adjustment of survivors of pediatric SCT, based on child, mother, and teacher reports, and identifies subgroups at risk for behavioral, social, and educational problems. Guided by the conceptual model adapted from Wallander and Varni's model,22 we also identified specific risk factors (eg, diagnosis of ALL, presence of GVHD, history of cranial radiation, child's poor health and young age, maternal distress) and resistance factors (eg, maternal maturity and higher education) for behavioral, social, and educational outcomes. Clearly, multiple perspectives and factors must be considered for identifying at-risk SCT survivors and families. This information can guide future research to focus on developing psychosocial and educational interventions to ensure optimal developmental trajectories for these patients and their families.
We thank Karen Sumbler, Tara Smith, and Gail Andrews for their help in collecting data for this study, and Fiona Schulte for her comments on an early version of this article.
This research was supported by grants from the Hospital for Sick Children Foundation, New Initiatives; Elizabeth Lue Bone Marrow Foundation; and the National Cancer Institute of Canada.