Discussion
The results of this multicentric study show that greater social
deprivation and therefore poverty, is associated with learning
difficulties in survivors after cancer treatment.
Learning difficulties were reported in 22% of the children in our
study, but higher rates have been noted in the literature(21,24).
Survivors are at greater risk for presenting learning difficulties with
scholarly consequences(10,15,21,24), ultimately leading to a lower
overall quality of life(25). School absenteeism induced by cancer care
is the most commonly reported productivity loss(8,26). Even if we did
not evaluate the rate of absenteeism in our cohort, we found that tumor
relapse was a risk factor. The tumor relapse frequently requires
extended hospital stays and therefore, induces school absenteeism.
In our study, the probability of declaring a learning difficulty at
school was associated with greater social deprivation. Irrespective of
the disease, there is a strong relationship between SES and scholarly
achievement(27–29). This supports our results and those reported in the
literature, even though different methods were used to estimate the SES
across studies(14,15,25). Parental involvement should also be considered
as a factor influencing academic achievement(30). Children with CNS
tumors are the population of greatest concern. CNS tumors and CNS
directed therapies (i.e., cranial radiotherapy and intrathecal
chemotherapy) increase the risk of adverse psychosocial
consequences(7,24,31). Academic and neurocognitive difficulties are
described, as well as higher rates of school absenteeism(8,9,21). We did
not confirm the inherent risk of radiotherapy in our CNS subgroup
analysis, probably due to the lack of children receiving irradiation
followed in our study. In this CNS subgroup, social deprivation remained
a prognostic factor for learning difficulties, an observation also shown
by Ach et al.(32).
HSCT was a risk factor for learning disabilities, in agreement with
previous reports that a high number of children receiving HSCT present
with academic difficulties and do not graduate high school(33).
Learning difficulties may be induced by a lack integration into the
school environment, as adequate socialization is a contributing factor
to academic success(29). Our data collection precludes this analysis;
however, Duan et al. demonstrated that greater social deprivation
negatively influences the relationship between academic socialization
and academic achievement(29). Survivors suffer from poor social
integration, potentiated by absenteeism(21,24,26,34). These difficulties
are described more frequently in patients with CNS tumors.
Due to the high risk for learning disabilities in survivors,
school-organized educational support remains necessary. Thirty-six
percent of children received a such support in our study, a rate
slightly higher than that reported in the literature, which ranged from
20% to 32.5%(24,31,35). Although greater deprivation was
associated with learning disabilities, it was not associated with the
probability of benefiting from academic support. Academic support seems
to be equally shared between children according to the EDI score,
whereas we showed a higher need in deprived areas.
Our analysis of psychological difficulties did not reveal an association
with social deprivation. However, SES may participate in causal
mechanisms for psychological problems (16,19). One study revealed that
neighborhood SES affected the probability of benefiting from
psychological support(36). Children with CNS tumors are at greater risk
for psychological effects, whereas our results indicated that children
with bone tumors were at risk(11,19). The psychological effects of bone
cancer can be explained by physical damage secondary to the surgery,
inducing vulnerability related to physical performance limitations,
disruption to routine activities, and diminished ability to attend work
or school(37,38).
This study is based on a prospective cohort and allowed us to evaluate
the overall psychosocial status of survivors after hospital treatment
and the evolution of psychosocial effects over time. However, this study
presents some limitations.
Our population was based on a database which aims to represent all the
children treated for cancer in the GOCE departments. However, one center
was excluded from the study because it did not participate to patients’
psychosocial evaluation. This excluded 6% of the children of the cohort
and therefore, we cannot exclude a selection bias (Figure 1).
Evaluation biases were present as a result of the variables’
measurement. Each outcome was evaluated during a consultation with the
oncologist pediatrician at a time when the clinical evaluation remained
the main objective to ensure remission persistence. The outcomes were
self-reported by the parents and children, and the subjective aspect of
self-reports cannot be ruled out. In addition, the data was only
collected at the posttreatment period and no baseline at diagnosis was
available. This precludes any analysis to evaluate the evolution of
psychosocial difficulties before and after the cancer management. This
limits us to an observation of post-treatment difficulties without
taking account of some potential predisposed conditions. Indeed, in
children with CNS tumors, preexisting predisposition syndromes (e.g.,
type one fibromatosis) can induce learning and psychological effects, as
well as the tumor symptoms themselves (39–41). However, the evaluation
of psychosocial status at diagnosis may be difficult. This time is a
period when clinical management remains the absolute priority. This can
explain the choice of investigators to defer the psychosocial evaluation
to after treatment.
The measurement of social deprivation using the EDI score assesses the
SES of the environment of each patient and can induce an ecological
bias. This measurement considers homogeneity between children living in
the same IRIS and could induce misclassification and underestimate the
effect of SES. Data on individual deprivation could be a complementary
method to precise the SES of children, but were not available in the
database.
Our statistical analysis is another limitation. Each child had a
different follow-up duration, which limited our utilization of the Cox
model. These different follow-up times resulted in missing data
organized in a monotone structure, resulting from dropout, and limited
the use of multiple imputations to consider the missing data.
Additionally, multiple imputation is not recommended for longitudinal
data analysis because of the lack of obvious benefit in this
context(42).