What Can Attention Abilities Teach Us about Reading Comprehension in NF1?

What Can Attention Abilities Teach Us about Reading Comprehension in NF1? Maëlle Biotteau1*, Elodie Tournay2, Eloise Baudou1,3, Sandrine Lelong4, Stéphanie Iannuzzi3, Nathalie FaureMarie3, Pierre Castelnau5,6,7, Elisabeth Schweitzer7, Diana Rodriguez8,9,10, Isabelle Kemlin8, Nathalie Dorison8, François Rivier11,12, Maryline Carneiro11, Elodie Preclaire12, Sebastien Barbarot13, Valérie Lauwers-Cancès2, Yves Chaix1,3 1ToNIC, Toulouse NeuroImaging Center, University of Toulouse, Inserm, UPS, France 2Epidemiology Department, Toulouse University Hospital, Toulouse, France 3Children’s Hospital, Toulouse-Purpan University Hospital, Toulouse, France 4Pediatric Clinical Research Unit, Toulouse Clinical Investigation Center, Children’s Hospital, Purpan University Hospital, Toulouse, France 5Brain & Imaging Joint Research Unit (UMR 930), Bretonneau Hospital, Tours Regional University Hospital, Tours, France 6University of Tours François Rabelais, Tours, France 7Neuropediatrics & Disabilities Unit, Gatien de Clocheville Children’s Hospital, Tours University Hospital, Tours, France 8Pediatric Neurology Department & Neurofibromatosis Referral Center, Armand Trousseau Hospital, East Paris University Hospital, Paris, France 9University of Paris VI Pierre & Marie Curie, Sorbonne Universities, Paris, France 10Neuroprotection of the Developing Brain Joint Research Unit (U1141), INSERM, Paris, France 11Department of Pediatric Neurology, CHU Montpellier, PhyMedExp, University of Montpellier, INSERM, CNRS, Montpellier, France 12Department of Pediatric Neurology & Reference Center for language disabilities, CHU Montpellier, France 13Dermatology Clinic, Hôtel-Dieu University Hospital, Nantes, France Research Open Access Journal of Neurophysiology and Neurological Disorders


Introduction Prevalence & Physical Characteristics of NF1
Neurofibromatosis type 1 (NF1) is one of the most common childhood genetic disorders, affecting approximately 1 in 2500 to 3000 individuals in the general population [1]. The diagnosis of NF1 is made based on physical characteristics as stated in the 1988 National Institutes of Health Consensus Development Conference Statement, and its 1997 update [2], which includes cutaneous, ophthalmologic, neurologic and orthopedic features [3].

Cognitive Deficits
A broad spectrum of cognitive deficits occurs in 30-70% of cases [4]. Studies focusing on the neuropsychological phenotype of NF1 children detect a left shift in average IQ, ranging from low to normal IQs [5], and specific deficits in several domains: attention processes, executive function, language acquisition, visuospatial and visual perception [6], and phonology [4,7,8], with the severity of these deficits varying greatly from one child to the next. NF1 cognitive problems also have the greatest impact on the quality of life [9,10]. They significantly impact academic performance [11,12] in particular reading [13] but also arithmetic and written expression. But, the specific contribution of attention disorders to academic difficulties has to date not been investigated.

Potential Causes of NF1 Reading Impairment
Studies of reading skills in NF1 have predominantly investigated discrepancies in IQ-achievements [4,14,15].
When compared to age-matched controls, Cutting et al. (2000) found that NF1 children had reading deficits which specifically affected (i) single-word reading and reading comprehension, (ii) letter-word identification and (iii) passage comprehension (as well as other skills associated with reading achievements such as rapid naming and phoneme segmentation). It has been estimated that between 50% [16] and 67% [15] of NF1 children suffer from reading deficits. Watts et al. (2008) further specified that 50% of NFI children satisfy the phonological dyslexia criteria and 13.3% of the mixed dyslexia criteria.
Reading is a multifaceted skill consisting of two primary components: word recognition and comprehension, with only one of these components, or both components impaired.
Different cognitive deficits may contribute to reading difficulties, specifically oral comprehension, working memory, phonological skills [8], visual perception [14], but also attention processes.
Considered one of the primary cognitive concerns in NF1 children [17,18,19], attention skills are a major contributor to reading impairment.

Attention Skills and Reading in ADHD Children and the General Population
It is now well-documented that elevated levels of inattention and/or hyperactivity/impulsivity are associated with various difficulties in the academic domain, including reading ability. This association has not only been confirmed in general populations of preschoolers [20], school-aged children [21,22] but also in adults [23]. Children with impaired attention, but who do not satisfy the attention deficit hyperactivity disorder (ADHD) criteria per se, have also been found to have reading problems [24,25]. It is widely accepted that there is a reciprocal causal relationship between attention and reading, where difficulties in one area intensify difficulties in the other [26].
Comorbid reading difficulties are also estimated to affect between 15 to 30% of ADHD children [27]. Indeed, ADHD children obtain lower reading comprehension scores compared to their peers without ADHD, which extends to all individual score parameters, such as the ability to read words, word identification, reading speed, vocabulary, cognitive skills, and background knowledge [28,29].
Furthermore, children with attention difficulties have lower reading comprehension scores, more difficulty reporting the central idea of a passage and their reading comprehension decreases proportionally with the increase in reading time length -potentially due to the attentional demands of reading lengthy passages which require more processing efforts [28].

Attention Skills and Reading in NF1 Children
In contrast to ADHD, very little is known about any potential associations between inattentive and hyperactive/impulsive behaviors and reading comprehension-related skills in the NF1 population. Even though 40% to 50% of NF1 children satisfy the ADHD diagnostic criteria [4,7,30,31,32], making ADHD one of the principal features of the clinical neurofibromatosis phenotype, the impact of attention skills on cognitive functions of NF1 children has received little coverage. We do however know that NF1 children with ADHD seem to experience more difficulties in terms of IQ, social skills, and some executive functions when compared to NF1 children without ADHD [31,32,33]. But visuospatial deficit [6,33], phonological deficit [8], and some execu-tive function impairments [4,32,34] appear to be independent of limitations in attentional abilities.
The literature to date is also vague about how ADHD affects cognitive functions associated with academic achievements. Although ADHD may affect the child's ability to learn, only one study has specifically investigated the association between attention abilities and reading in an NF1 population [35].

The Complexities of Assessing Attention Skills
The assessment of attention skills poses substantial challenges. Firstly, there is a large diversity of attention domains that can be assessed using multiple approaches. These approaches can nevertheless be broadly subdivided into two major classes: "indirect observer-rated data" (questionnaires, checklists, interviews, think-aloud protocols, global rating scales, self-report instruments, direct observation of behavior, etc.) and "direct performance-based data" (performance-based measures of functional skills, indices of competence, real-world outcomes).
"Indirect observer-rated data" provides a generalized picture of attention performance across everyday functions, in line with the diagnostic criteria defined by the DSM. This type of data is completed by individuals who are expected to make a judgment on a child's behavior (i.e. teachers, parents, etc.). They are the most commonly used forms of assessment (particularly rating scales), predominantly because of their efficiency. Indirect observer-rated data, however, tend to be abstract and introduces the subjective perspective of the respondent (often parents or teachers), ranging from different expectations across environments to inter-rater disagreements, bias, and different behavioral tolerances [36,37]. Rating scales, for example, often have a poor teacher and parent inter-rater agreement [38,39]. The extrapolation of results based on indirect observer-rated data therefore requires considerable vigilance. "Direct performance-based data" on the other hand, is collected directly from the children.
This approach may provide a more direct and valid estimate of attention abilities. This type of data has the additional benefit of circumventing response bias or can be adapted to capture specific biases. However, performance-based measures of attentional capacity require the subject to actively participate in the assessment process (motivation and cooperation), which is not always evident in practice, especially in children. Moreover, they present an artificial evaluation as they don't capture the child's performance in everyday life settings (the validity of these tests in different environments remains unclear).
In addition, attention is a continuum and should be measured as such. Particularly when identifying pathological vs non-pathological degrees of attention impairment (ADHD vs. non-ADHD) or addressing questions regarding diagnosis or frequency of co-occurrence between two pathologies. Assessments of attention therefore need to take into account that attention develops along a continuum. Inattention, hyperactivity and impulsivity have been consistently proven to be associated with specific domains of daily function or learning, even in children without elevated behavioral problems, and therefore require the child's behavior to be assessed dimensionally [40].

Aims and Objectives
Reading and attentional impairments occur together in

Procedure
Participants underwent neuropsychological assessments, using a comprehensive and extensive protocol designed to assess the cognitive level, reading skills, phonological process, visuoperceptual abilities and attention (procedure previously described by Chaix et al. [8].

Measures
(i) The cognitive assessment included all subtests of the Wechsler Intelligence Scale for Children -Fourth Edition [42]; and the French version of the Peabody Picture Vocabulary Test-Revised (EVIP) [43].
(ii) Reading and phonological skills were assessed with the "L' Alouette" French reading test [41], the ODEDYS-2 test [44] and the ORLEC battery subtest L1 and L3 [45]. The reliability and validity of these three tests are deemed satisfactory for most measurements [41,44,45]. The "L'Alouette" test assesses the level of lexical decoding, two indices of accuracy and speed when reading a text aloud. These indices were standardized for age to obtain z-scores. Reading accuracy and speed disabilities were defined as a score of at least -1.5SD, below the normative mean value for children of similar ages. The ODEDYS-2 test measured word recognition processes on a series of 20 regular words, 20 irregular words and 20 non-words (pseudo-words), to be read aloud. Both accuracy and speed were considered. This test was used to subdivide the reading profile of participants (e.g., phonological, surface, or mixed dyslexia). The ORLEC battery assessed reading comprehension of sentences as well as texts [45]. The first subtest (L1) consisted of a text to be read aloud. Children (iii) The visuo-perceptual assessment was performed using the Judgment of Line Orientation test [46], the Thurstone test and the Corsi blocks.
(iv) Attention skills were assessed with the Conners' Continuous Performance Test-Second Edition (CPT-II) [47] and the parent form of the Child Behavior Checklist (CBCL) questionnaire [48].

The Conners' Continuous Performance Test-Second Edition
(CPT-II) [47] computerized test is designed to specifically measure sustained attention and impulsivity. It is generally used for differentiating children with and without inattentive, hyperactive, and impulsive behavioral difficulties [49,50]. The test is 14 minutes long and during this time, the child must refrain from pressing the space bar whenever the letter "X" appears but is required to press the space bar for any other letter (target/non-target stimuli). CPT-II is designed to have minimal language and memory demands. Three percentile scores were used in this analysis: the omissions score (number of non responses to target), the commission's score (number of responses to non-target stimuli) and the Hit RT Standard Error score (a measure of response speed consistency). High percentile scores reflect significant attention problems and the disability level for each dimension was defined as a percentile score above or equal to 90. The parental form of the Child Behavior Checklist (CBCL) [48] questionnaire measure attention problems and psychosocial adjustment. The CBCL "attention problems" subscale which focuses on 11 questions among the 113 in the complete CBCL test was used in this analysis. First, a raw-score was computed as the sum of the 11 scores between 0 and 2, then standardized to obtain the z-score which expresses the number of standard deviations (SD) from the mean in the general population of the same gender and age.
Raw-scores ≥ 11 were considered to indicate "attention problem" disabilities.

Statistical Analysis
Attentional capacities, reading skills, IQ, and sociocultural characteristics were compared between NF1 and non-NF1 children using the paired Student T-test and the Wilcoxon signed-rank test (when the normality assumption of differences NF1 minus control was rejected) for continuous variables, the McNemar test for binary variables, and the Bowker's test of symmetry for categorical variables with more than 2 categories.
Spearman correlation coefficients were computed to assess the correlations between the CBCL "attention problem" z-score and the CPT-II percentile scores separately, in NF1 and in non-NF1 children, and then on all children with disabilities in CBCL. Significance levels (p-values) were used to assess whether the correlation coefficients were significantly different from zero.
To identify which factors were associated with the text and sentence comprehension scores (Lobrot), a multivariate analysis of covariance was applied to all children. Regressors initially included in the models were CBCL "attention problem" z-scores, omission, commission percentile scores (CPT-II), EVIP normalized score, IQ and reading speed and accuracy z-scores (L' Alouette). The Hit RT Standard Error percentile score was not used in this multivariate analysis due to its high correlation with omissions which were considered more relevant clinically.
Classification variables were the population (NF1 vs. non-NF1) as well as the parents' educational and socio-professional levels. Only

Attentional Capacity
The measures of attentional ability are presented in

Reading Skills
As shown in Table 1, text and sentence comprehension (Lobrot) were not significantly different between NF1 and controls. While

Correlations between CBCL and CPT-II Attention Tests
A weak correlation, between CBCL "attention problem" 6

Modeling of Text and Sentence Comprehension
Models to explain text and sentence comprehension initially included the NF1 and control groups, the CBCL "attention problem" z-scores, omission, commission percentile scores of CPT-II, IQ, reading speed and accuracy z-score (L'Alouette), EVIP normalized score, and parents' socio-professional category and education level. The final models obtained after the elimination of variables are presented in Table 3.

Discussion
The aim of our study was to investigate the association between attention problems and reading comprehension ability

Neurofibromatosis Type 1 Challenges How Attention is Measured?
We used an indirect parental assessment (checklist, CBCL) and a performance-based measure (CPT-II) of attention to determine whether attention processes affect reading skills.
There was no significant concordance between these two assessment approaches, in the NF1 group and also not in the group of children without NF1. This could either be explained by a lack of relevance of at least one of these assessment tools or alternatively that the parental questionnaire and neuropsychological tests evaluate different dimensions. Direct behavioral performance-based data are more reliable than rating scales but are expensive, time-consuming, do not take into account the child's everyday environment [34], and can be difficult to administer to children as they rely on the child's willingness to cooperate.

Direct and Indirect Attention Assessment Tools: An Overlap?
Although a few studies have performed both direct and indi-

What about Direct and Indirect Attention Assessment in NF1
Specifically?
Our findings are consistent with previous NF1 results that used neuropsychological tests in conjunction with functional questionnaire measurements [59]. Payne et al. (2011) [60] which found that questionnaires tap, more specifically, into real-world functions that are quite different from the cognitive assessment of the same function under standardized conditions (i.e. a quiet room) and highlight the importance of taking these two dimensions into account to guide remediation programs. In conclusion, no single test will capture all aspects of the complex attention domain in NF1 children, the preferable approach is therefore to combine performance-based and observer-rated outcome measures in a multimodal assessment of attention processes [62].

Underlying Mechanisms of Attention on Reading Comprehension in NF1
No Correlation, no Link, and Not Even the Same Processes: We detected a significant association between attention variables and reading comprehension when scored with direct and indirect tools, and in children with or without NF1. The contribution of attention components in reading-related activities has been previously demonstrated [63]. Children with ADHD may be more likely to show deficits in reading and other academic skills [64].
A study of school-aged children [65] has also revealed correla-

What about the Implication of Attention in Reading Comprehension in NF1 Specifically?
Attention is one of the primary cognitive concerns for NF1 children [17,18,19]. Although ADHD may affect the child's ability to learn [29], attention processes have received little consideration from the research community, to explain the reading deficit. Only one study [35] has investigated and showed a link between attention and academic achievement including reading, in NF1 (albeit by examining academic achievement rather than reading specifically). This reading score only assessed the reading of single words (Word Reading Spelling subtest of the WIAT) and therefore failed to score the complexity of the reading process. Furthermore, the precise nature of the association between reading and attention remains vague. In particular, the direction of causation cannot be inferred by the results obtained and is indistinguishable if Pride's results indicate that the presence of ADHD increases the presence of learning disabilities or vice versa. Indeed, the ADHD prevalence rate increased in NF1 children when other disorders such as Specific Learning Disabilities coexisted [4].
Overall, the literature to date is quite unclear on how ADHD affects cognitive functions in NF1. Furthermore, we believe that the precise nature of the association between attention and cognitive functions has not been adequately (and appropriately) studied. Indeed, all studies that have taken an interest in this question focused solely on evaluating NF1 alone or NF1 associated with ADHD (i.e. pathologic vs. non-pathologic ). As attention abilities are a continuum in NF1 (and all other populations), the best means to study attention problems and to determine their impact on academic achievement, is to analyze attention as a continuous variable rather than comparing groups with and without ADHD [59]. In accordance with the Lehtonen recommendation, our finding demonstrates the impact of attention level on reading comprehension level (and not only on ADHD).
Our study therefore provides a strong argument supporting that attention difficulty in the NF1 child will have an impact on the child's academic performance and, particularly, on learning to read.

Capturing Highly Specific Reading Skills: What Attention Tests Need to be Used?
To determine which were the greatest "predictors" of reading comprehension, we used models to explain text and sentence When the measure of reading accuracy increases, the measure of reading comprehension increases (Lobrot score), meaning that improvements in the child's reading accuracy, are reflected in an improvement in reading comprehension. To summarize, a good understanding of both sentences and texts is associated with a high IQ, good reading accuracy and few inattention problems.
Reading comprehension is a complex process that requires the coordination of numerous cognitive abilities (in order to understand words, sentences, and texts), including phonological and orthographic knowledge, correspondence to letters, decoding, memory, etc., but also the general cognitive ability (IQ). Numerous studies have shown that reading is correlated with language abilities (verbal IQ), phonological awareness, working memory, and attention [64,68]. For IQ in particular, several studies have investigated the proportion of variance in reading ability that is explained by IQ [69]. However, even if an association is often evoked [68,69,70], reading has an etiology that is not completely dependent on shared influences with IQ [69]. Furthermore, IQ is not often used in practice as a predictor of reading ability due to the existence of children with average reading ability and low IQ or average IQ and low reading ability (as in dyslexia). This implies that factors other than IQ are critical in the development of successful reading abilities, which is confirmed by our current study. Indeed, we report the exact nature, variance, and overlap between the contribution of components that affect reading comprehension. Our study therefore supports that reading is a multifaceted skill set and a complex process that encompasses reading accuracy, IQ and attention, and requires the implementation of several tools and evaluation approaches.

Difference between NF1 and non-NF1 Individuals:
We showed that lower reading speeds are associated with a decrease in reading comprehension in both NF1 and non-NF1 populations, but that this association was stronger in the NF1 group. This may indicate that the effect of reading speed on reading comprehension is modulated by a third, as yet not identified factor, in the NF1 population.

The Specific Role of Attention in Reading Comprehension:
We found that attention level, rather than impulsivity or hyperactivity, is an effective predictor of the reading comprehension level in NF1 children as well as in non-NF1 children. This result is in line with previous studies [20,66,71] that revealed that inattentive behaviors were strongly inversely associated with reading (particularly fluency and comprehension), but hyperactivity and impulsive behaviors did not predict reading achievement.
However, these studies did not consider reading and attention as continuums and only considered the tail end of the normal curve (RD and ADHD as pathological conditions). They did not investigate the normal variation in the reading, attention interaction. In this sense, our findings confirm those obtained in earlier studies, showing the specific effects of attention on reading achievement when compared to impulsivity or hyperactivity. But they also improve on them, by highlighting for the first time, the definite role of attention in reading comprehension, even outside of its pathological forms.
We also surprisingly detected a correlation between reading comprehension and impulsivity. Indeed, we found that when the measure of impulsivity increased (CPT-II percentile commission), the measure of sentence reading comprehension decreased (Lobrot score). Here, the CPT-II commission had an effect on sentence reading comprehension (not significantly different between the 2 populations, NF1 and non-NF1). For example, an increase of 10 points in the commission score (impulsivity) led to an increase of 0.5 points in the Lobrot score (higher understanding). This result is not congruent with clinical reality.
Firstly, children who were less impulsive presented a better analysis of the sentence or text and consequently a better comprehension. Secondly, it is most likely inattention rather than the hyperactivity that forms the association between ADHD and reading [22]. We therefore consider this result to be an artifact. commission's score, in this case, appears to be a better marker of response time, than a measure of real impulsivity. We can also discuss the results of a recent study in which authors reported that impulsive behavior was positively related to reading skills [72]. More precisely in their study, individuals with a medium to a high level of hyperactivity and impulsivity but with a low level of inattention tended to perform better on reading measures than individuals with low-level hyperactivity and impulsivity but a medium level of inattention.
To summarize, (1) good sentence and text reading comprehension are not only associated with attentional measures but also with higher IQ, and good reading accuracy, irrespective of the population studied; (2) reading speed is also associated with good sentence and text comprehension in non-NF1 children,

Conclusion
In the present report, we have focused on reading comprehension, attention assessment tools, in NF1 and non-NF1 children. When attention problems in NF1 children are analyzed as a continuum as opposed to two groups (ADHD and not ADHD), we found that inattention, rather than impulsivi- There is an overall lack of consensus regarding how best to measure attention, and also whether both types of assessments (direct test-based and indirect questionnaire-based) may be used for the same purposes, especially in the context of investigating correlations with other cognitive measurements. We showed that both indirect and direct assessments are complementary to evaluate attention skills as they explore different domains. They are not interchangeable and should be used in parallel. The finding that these individual measurements demonstrate a unique association with a specific reading skill implies that both should be considered in the diagnosis process.
Moreover, the development of neuropsychological tests that are more sensitive to the child's test environment should be prioritized, particularly for evaluating NF1 children.
Acknowledgments: This work is dedicated to the children and families who agreed to participate in the study. We would like to thank the NF network France, ANR France, and the Occitanie Region for their support. The authors also thank Laurence Lion-François for her help and Petra Neufing for the English language editing. Children with or without NF1 (n=150; 8-12 years of age), matched for age, gender, handedness, and reading level were submitted to direct and indirect attention tools as well as text and sentence reading comprehension. For both NF1 as well as non-NF1 children, attention capabilities greatly influenced 13 reading comprehension, but there was only a weak correlation between direct and indirect attention measurements. Indirect observer-rated (questionnaires) and direct performance-based assessments of attention measure different components of reading skills. We make the case that children with NF1 would benefit from a multimodal assessment of attention skills, particularly where cognitive functions are evaluated to detect causal associations.