The clinical sample was recruited from a referral center for NF1 located in a university hospital. Participation was voluntary and offered to all parents during medical consultations carried out between May 2013 and March 2016. To be included, children had to meet the diagnostic criteria of the National Institutes of Health Consensus Conference1 and be aged between 8 and 12 years. Learning disabilities were not an inclusion or exclusion criterion, to avoid recruitment bias13. The exclusion criteria were epilepsy or brain tumor in the context of NF1, history of another neurological or psychiatric pathology, sensory disorder inconsistent with testing (especially symptomatic optic chiasm glioma), and insufficient French language proficiency. All 52 families to whom the study was proposed agreed to participate. One child was excluded because of a prior history of head trauma, and 11 others were not included in the analyses, owing to too many missing data. The final sample therefore included 40 children with NF1.
Healthy control children were matched with the clinical sample for age, sex, and mean parental education level, as the standardized psychometric tests available to assess EFs in France have not all been validated with French children. We recruited 71 control children through different networks (schools, holiday, and leisure centers). The inclusion criteria were the same as for the clinical sample, apart from the fact that they were not expected to have NF1 or a history of learning disabilities. We used the Full Scale Intellectual Quotient (M = 100, SD = 15) derived from the four primary indices of the French Wechsler Intelligence Scale for Children‑4th Edition14 to exclude intellectual disability. A total of 12 children were not retained, owing to high number of missing data, and three others were removed from the study because of suspected attention deficit hyperactivity disorder. The final control sample consisted of 56 children.
General health-related QoL was measured with the validated French-language version of Kidscreen-5215. Children (self-report) and their parents (proxy report) responded to 52 questions covering 10 QoL domains (Physical wellbeing, Psychological wellbeing, Moods and emotions, Self-perception, Autonomy, Parent relations and home life, Social support and peers, School environment, Social acceptance, Financial resources) on a 5-point scale. The raw scores (M = 50, SD = 10) were converted into T scores.
To take the current recommendations for EF assessment16 into account, we used both direct (i.e., psychometric tests based on the child’s actual performance in the examination setting) and indirect (behavioral inventory) measures. We combined seven performance-based tests with two behavioral inventories targeting daily life. Three of the performance-based tests (i.e. Rey-Osterrieth complex figure (ROCF), Stroop test, and Modified Wisconsin Card Sorting Test) do not have a validated French-language version, but were selected because of their clinical sensitivity in NF16, 17, 18. For these tests, we used raw scores. Four other tests were administered, namely the Two Barrages Test (T2B)19, the Digit span and Letter‑number sequencing subtests of the Wechsler Intelligence Scale for Children14, and the Auditory attention and response set from the Developmental Neuropsychological Assessment 20. We used Z scores for T2B and standardized scores for the Developmental Neuropsychological Assessment (M = 10, SD = 3). For the Digit span and Letter-number sequencing subtests, the scores were pooled to obtain a composite working memory index (WMI; M = 100, SD = 10), in accordance with the Wechsler manual's recommendations. The French validation21, 22 of the Behavior Rating Inventory of Executive Function (BRIEF23) was completed by parents and teachers, and T scores were used to calculate the three composite indices: behavioral regulation (BRI), metacognition (MI), and global executive composite.
Disease severity was assessed with the revised Riccardi scale24, ranging from 1 (Minimal) to 4 (Severe). Disease visibility was assessed with the Ablon scale25, ranging from 1 (Mild) to 3 (Severe).
The extent of school and extracurricular support received was assumed to provide an indirect measure of learning disabilities. School support corresponded to the educational services put in place at school (i.e., number of months in which the child received help such as a school-life support, care service, school support, personalized schooling, or personalized educational success program). Extracurricular support corresponded to out-of-school care (i.e., number of months the child received speech or motor therapy or psychological follow-up).
The study was approved by the institutional review board (CPP Est III, 12 March 2013, ID-RCB Number: 2012-A00787-36) and registered with the French Data Protection Authority (CNIL, EGY/VCS/AR135993). Written consent was signed by the child and at least one parent after they had read an information note. The testing protocol was carried out by two experienced psychologists, assisted by four Master's students.
The protocol was part of a larger study of neuropsychological disorders in NF1 (3 sessions each lasting 150 minutes). The children with NF1 were seen as part of their neuropsychological assessment at the hospital, while control children were seen at home. The tests were administered in a predefined and systematized order.
All analyses were carried out with R software (R Core Team, www.r-project.org). Differences between the NF1 and control groups on raw or standardized scores (based on calibrations) on the various measures were examined with two-tailed Student t tests. Welch's correction was applied to degrees of freedom to account for the heterogeneity of variances in the population. A value of p < .05 was considered significant and a value of p ≤ .10 as a trend. Effect sizes were calculated and analyzed according to Cohen's recommendations26: small if 0.2 ≤ d < 0.5; moderate if 0.5 ≤ d < 0.8; large if d ≥ 0.8.
Linear regressions were calculated to examine the determinants of low QoL domain scores (self- and proxy reports). Two kinds of predictors were considered and selected on the basis of a critical threshold of .10, namely (1) low EF scores, and (2) sociodemographic (age, sex, parental education level) and disease-related (sporadic vs. familial form, severity, visibility) factors. We also included the extent of school and extracurricular support (see 27, 28), in order to contrast it with other indicators. Any missing data for a regression were completed ("impute" or "fill in") with the Amelia 2 package 29, 30. The parameters of the function assumed that the data were multivariate normal. This program uses the EMB algorithm (B for bootstrap31). A total of 100 completed datasets were created (by default with Amelia 2) and backed up. Regression results (beta and SE) were pooled according to standard multiple imputation rules32.