INTRODUCTION
Neurofibromatosis type 1 (NF1) is a single-gene disorder that affects approximately 1 in 2700 people (Evans et al., Reference Evans, Howard, Giblin, Clancy, Spencer, Huson and Lalloo2010). Learning difficulties are one of the most frequent complications in childhood (Lehtonen, Howie, Trump, & Huson, Reference Lehtonen, Howie, Trump and Huson2013), with poor reading being one of the most significant challenges for children with NF1 (Cutting, Clements, Lightman, Yerby-Hammack, & Denckla, Reference Cutting, Clements, Lightman, Yerby-Hammack and Denckla2004). Although impairments in basic word reading (Watt, Shores, & North, Reference Watt, Shores and North2008) and reading comprehension (Cutting, Koth, & Denckla, Reference Cutting, Koth and Denckla2000) are consistently reported in children with NF1, the specific nature of these impairments and contributing cognitive and literacy factors are not well understood.
Reading aloud is complex, with a number of cognitive processes involved (Coltheart, Reference Coltheart2005). The dual-route cascaded (DRC) model is a well-established framework that theorizes the cognitive processes needed for accurate word reading (see Figure 1). According to the DRC model (Coltheart, Rastle, Perry, Langdon, & Ziegler, Reference Coltheart, Rastle, Perry, Langdon and Ziegler2001), when a child initially sees a written word (e.g., SHIP), they must first correctly identify the letters and their order in the word (letter identification). The output of this process triggers further processing in two routes. The sublexical route translates each grapheme (e.g., SH I P) into a phoneme (“sh” “i” “p”) by applying the child’s existing grapheme-to-phoneme correspondence (GPC) knowledge. At the same time, the lexical route carries out a “look-up” procedure whereby the visual form of the written word (e.g., SHIP) is searched for in their orthographic lexicon of all learned written words. If a match is found, this triggers the spoken version of the word (e.g., “ship”) in the phonological lexicon – sometimes via the semantic system if it comprises the meaning of the written word (e.g., a big boat). Both the sublexical and lexical routes link again at the phonological output system that contains information that enables the word to be pronounced aloud.
Fig. 1. Dual-route model of reading (Coltheart et al., Reference Coltheart, Rastle, Perry, Langdon and Ziegler2001). Adapted from Figure 1 (McArthur, Kohnen, et al., Reference McArthur, Kohnen, Larsen, Jones, Andakumar, Banales and Castles2013).
Children need to develop both the sublexical and lexical routes to become a skilled reader of the English language (Castles & Coltheart, Reference Castles and Coltheart1993) since written English includes regular and irregular words. However, not all children have fully functioning reading routes. Some children have an impaired sublexical route or “phonological dyslexia” (Castles & Coltheart, Reference Castles and Coltheart1993) characterized by poor reading of nonwords (i.e., nonsense words that can be read using GPC knowledge, e.g., gop). Other children have an impaired lexical route or “surface dyslexia,” characterized by impaired reading of irregular words (Castles & Coltheart, Reference Castles and Coltheart1993). Children with both routes impaired have “mixed dyslexia,” which occurs most frequently in children with poor word reading skills (Castles & Coltheart, Reference Castles and Coltheart1993; McArthur, Kohnen, et al., Reference McArthur, Kohnen, Larsen, Jones, Andakumar, Banales and Castles2013).
To date, one study that used the DRC model to investigate subtypes of dyslexia (Watt et al., Reference Watt, Shores and North2008) found that two-thirds of children with NF1 (20/30) demonstrated a significant reading impairment (i.e., performance in the bottom 5% for nonword and/or irregular word reading), with 75% of these children having phonological dyslexia, 20% having mixed dyslexia, and no children having surface dyslexia. This distribution of subtypes differs substantially from that of children in the general population, where using the same criteria, 17% of poor readers have surface dyslexia and 9% have phonological dyslexia (McArthur, Kohnen, et al., Reference McArthur, Kohnen, Larsen, Jones, Andakumar, Banales and Castles2013).
Children with NF1 commonly experience a range of cognitive and behavioral difficulties (Lehtonen et al., Reference Lehtonen, Howie, Trump and Huson2013), and it is not clear how these difficulties impact their reading abilities. Deficits in phonological awareness (i.e., detecting and manipulating sounds of language) and phonological memory (i.e., the ability to store and manipulate sounds in short-term memory) have been reported (Arnold, Payne, Lorenzo, North, & Barton, Reference Arnold, Payne, Lorenzo, North and Barton2018; Chaix et al., Reference Chaix, Lauwers-Cancès, Faure-Marie, Gentil, Lelong, Schweitzer and Castelnau2017; Cutting et al., Reference Cutting, Koth and Denckla2000; Mazzocco et al., Reference Mazzocco, Turner, Denckla, Hofman, Scanlon and Vellutino1995). In addition, the relationship between the Judgement of Line Orientation (JLO) (Benton, Sivan, Hamsher, Varney, & Spreen, Reference Benton, Sivan, Hamsher, Varney and Spreen1983), which is frequently impaired in children with NF1 (Lehtonen et al., Reference Lehtonen, Howie, Trump and Huson2013), and reading is not clear. One study found that children with NF1 and reading difficulties display distinct visuospatial deficits on the JLO when compared to children with reading difficulties and controls. In addition, only for children with NF1 and reading difficulties did their single-word reading skills correlate positively with JLO performance (Cutting & Levine, Reference Cutting and Levine2010). In the general population, children with reading disabilities also experience visuospatial deficits that are age dependent (Giovagnoli, Vicari, Tomassetti, & Menghini, Reference Giovagnoli, Vicari, Tomassetti and Menghini2016) and are proposed to be associated with visual magnocellular pathway abnormalities (Stein, Reference Stein2019). While Cutting and Levine (Reference Cutting and Levine2010) found no significant relationships between reading and cognitive factors (e.g. language, attention) in children with NF1, future studies are needed given the relatively small sample size (Cutting & Levine, Reference Cutting and Levine2010) and to examine other factors that may contribute to poor reading outcomes, such as working memory (Wang & Gathercole, Reference Wang and Gathercole2013), sex (Quinn & Wagner, Reference Quinn and Wagner2015), and socioeconomic status (SES; Hecht, Burgess, Torgersen, Wagner, & Rashotte, Reference Hecht, Burgess, Torgersen, Wagner and Rashotte2000). Developing a better understanding of these impairments and related factors is vital for the design of effective reading interventions.
Therefore, the aims of this study were to (1) identify the incidence of reading impairments in children with NF1 compared to an unaffected comparison group; (2) categorize the types of reading impairments in poor word readers with NF1 and compare the profile associated with each type; and (3) identify cognitive and demographic predictors of poor word reading in children with NF1. We predicted that (1) as a group, children with NF1 are at a greater risk of being a poor reader than typically developing (TD) controls; (2) based on Watt et al. (Reference Watt, Shores and North2008) study findings, the majority of poor readers with NF1 would be classified with phonological dyslexia, with fewer cases demonstrating mixed dyslexia and no cases of surface dyslexia; and (3) reading would be associated with gender, SES, working memory, attention, language, and visuospatial abilities.
METHOD
Sixty-two children, aged 7–12 years, who met clinical diagnostic criteria for NF1 (National Institutes of Health, 1988) were recruited from the Neurogenetics Clinic at The Children’s Hospital at Westmead in Sydney, Australia. Families were offered the opportunity to participate during routine evaluation of their child’s neuropsychological functioning. Participants were also recruited via an invitation letter mailed to families who attend the clinic with children within the study age range.
NF1 participants were excluded if they had (1) diagnosed intracranial pathology (i.e. symptomatic optic gliomas), (2) vision or hearing loss that may impact the validity of assessment, (3) inadequate English to complete tasks, or (4) an extremely low full-scale intelligence quotient (FSIQ < 70). Two children with NF1 were excluded following IQ assessment (FSIQ < 70), resulting in 60 children with NF1 in the final analysis.
An unaffected peer comparison group of 36 TD children were recruited from a longitudinal study being conducted at the hospital (Lorenzo, Barton, Arnold, & North, Reference Lorenzo, Barton, Arnold and North2015), advertisements placed within local primary schools, community newspapers, and the hospital. A structured parent interview regarding the child’s development was conducted prior to the assessment. No TD child had a reported history of genetic, neurological, or psychological disorders, intellectual impairment, learning difficulties, or developmental delay.
Approval for this study was granted by the Sydney Children’s Hospital Network (HREC/11/CHW/28) Human Research Ethics Committees, and the study was conducted in accordance with Helsinki Declaration. Written informed consent was obtained from all participants. All participants were individually assessed at the hospital by a psychologist with short breaks provided as needed. Parents completed questionnaires regarding their child’s development and behavior.
Measures
For those tests with normative data, raw scores were converted to standard scores using the published test manual or the methods and normative tables contained within the citation for the test. Unless specified otherwise, all scores were normed according to age, and higher scores indicate better performance.
Cognitive and demographic measures
IQ: General intellectual functioning was assessed with either the Wechsler Intelligence Scale for Children – Fourth Edition, Australian Adaptation (WISC-IV; Wechsler, Reference Wechsler2003) or the Wechsler Abbreviated Scale of Intelligence – Second Edition (WASI-II; Wechsler, Reference Wechsler1999). The WASI-II was used for control and NF1 participants (n = 16) who had not undergone an IQ assessment in the preceding 12 months. As the correlation coefficient between these two measures is very high for FSIQ (r = 0.86; Homack & Reynolds, Reference Homack and Reynolds2007), scores were pooled across participants to provide an overall estimate of intellectual functioning (FSIQ), verbal IQ (Verbal Comprehension Index (VCI)/Verbal IQ (VIQ)) and performance IQ (Perceptual Reasoning Index (PRI)/Performance IQ (PIQ)) (standard score; M = 100, SD = 15).
Verbal working memory: Children assessed with the WASI-II completed the Working Memory subtests of the WISC-IV (standard score; M = 100, SD = 15) (Wechsler, Reference Wechsler2003).
Attention: Children’s ability to focus, sustain, switch, and divide their attention was assessed by Sky Search, Score, Creature Counting, and Sky SearchDT from the Test of Everyday Attention for Children (scaled score; M = 10, SD = 3) (Manly, Robertson, Anderson, & Nimmo-Smith, Reference Manly, Robertson, Anderson and Nimmo-Smith1999). Parents also completed the Conners-3 (Conners, Reference Conners2009), yielding T scores for the Inattention and Hyperactivity/Impulsivity scales (standard score; M = 50, SD = 10). Higher T scores indicate more problem behaviors.
Visuospatial functioning: The JLO task (Benton et al., Reference Benton, Sivan, Hamsher, Varney and Spreen1983) was used to assess visuospatial abilities as children with NF1 typically display impairments on this measure (Clements-Stephens, Rimrodt, Gaur, & Cutting, Reference Clements-Stephens, Rimrodt, Gaur and Cutting2008). Raw scores were converted to z scores (M = 0, SD = 1) using published normative data.
SES: SES was estimated using the Index of Relative Socio-Economic Advantage and Disadvantage, which ranks geographic areas in terms of their socioeconomic advantage and disadvantage (“Statistical Local Area, Table 2: Socio-Economic Indexes for Areas (SEIFA), data cube: Excel spreadsheet, cat. no. 2033.0.55.001,” 2011). The lowest 10% of areas are ranked a decile of 1 and the highest 10% are ranked a decile of 10.
Reading measures
Single-word reading: Single-word reading abilities were assessed with the Castles and Coltheart Word/Nonword test – Second Edition (CC2; Castles et al., Reference Castles, Coltheart, Larsen, Jones, Saunders and McArthur2009), a revision of the original Word/Nonword test (Castles & Coltheart, Reference Castles and Coltheart1993). The revised test has an expanded set of words to minimize ceiling effects, incorporation of a stopping rule, changes in the order of word presentation, and updated norms. Raw scores for regular, irregular, and nonword reading scores were converted to age-adjusted normalized z scores (M = 0, SD = 1) using the normative tables reported by Castles et al. (Reference Castles, Coltheart, Larsen, Jones, Saunders and McArthur2009) that were calculated from a sample of 1036 Australian children (6–12 years old).
Word reading fluency: The Test of Word Reading Efficiency – Form A (Torgesen, Wagner, & Rashotte, Reference Torgesen, Wagner and Rashotte1999) was administered to provide a measure of sight word and phonemic decoding efficiency (standard score; M = 100, SD = 15).
Reading comprehension: Reading comprehension was assessed by the Test of Everyday Reading Comprehension (McArthur, Jones, et al., Reference McArthur, Jones, Andakumar, Larsen, Castles and Coltheart2013). Raw scores were converted to z scores (M = 0, SD = 1) based on the child’s age using normative tables reported by McArthur, Jones, et al. (Reference McArthur, Jones, Andakumar, Larsen, Castles and Coltheart2013) that were calculated from a sample of 535 Australian children (6–12 years old).
Common processes (lexical and sublexical)
Letter identification: The cross-case copying measure was used to assess children’s basic letter knowledge (McArthur, Coltheart, Castles, & Kohnen, Reference McArthur, Coltheart, Castles and Kohnen2008) and consisted of 14 items. For the Letter Orientation Task (LOT), children were asked to identify whether letters were facing the correct way or “flipped backwards” (McArthur et al., Reference McArthur, Coltheart, Castles and Kohnen2008). For both measures, each correctly answered item was awarded one point. Raw scores are reported.
Phonological output: Children’s phonological output was assessed by (i) the Repetition of Nonsense Words subtest from A Developmental Neuropsychology Assessment – Second Edition (scaled score; M = 10, SD = 3) (Korkman, Kirk, & Kemp, Reference Korkman, Kirk and Kemp2007) and (ii) the Blending Nonwords test (McArthur et al., Reference McArthur, Coltheart, Castles and Kohnen2008) which contains 28 items, with higher scores indicating better performance. Each correctly answered item was awarded one point. Raw scores are reported.
Lexical processes
Orthographic lexicon: The Test of Orthographic Choice (Kohnen, Anandakumar, McArthur, & Castles, Reference Kohnen, Anandakumar, McArthur and Castles2012) assessed children’s written word recognition. Raw scores were converted to z scores (M = 0, SD = 1) using grade-based normative tables that were calculated from a sample of 102 Australian children (Grades 1–6) (Kohnen et al., Reference Kohnen, Anandakumar, McArthur and Castles2012).
Phonological lexicon: The Rain/Hane task assessed children’s phonological representation of 20 spoken words (McArthur, Kohnen, et al., Reference McArthur, Kohnen, Larsen, Jones, Andakumar, Banales and Castles2013). Each correctly answered item was awarded one point. Raw scores are reported.
Semantic knowledge: To assess semantic knowledge, the Peabody Picture Vocabulary Test – Fourth Edition (Dunn & Dunn, Reference Dunn and Dunn2007) was administered (standard score; M = 100, SD = 15).
Sublexical processes
GPC knowledge: To assess letter–sound knowledge, the Letter–Sound Test comprising 51 items (Larsen, Kohnen, Nickels, & McArthur, Reference Larsen, Kohnen, Nickels and McArthur2015) and the Grapheme-Phoneme Correspondence (GPC) test (McArthur et al., Reference McArthur, Kohnen, Jones, Eve, Banales, Larsen and Castles2015) consisting of 39 items were administered. For both measures, each correctly answered item was awarded one point. Raw scores are reported.
Categorizing Word Reading Impairments
Children were categorized as poor readers if their z score for nonword and/or irregular word reading on the CC2 was at least 1 standard deviation below the mean (McArthur, Kohnen, et al., Reference McArthur, Kohnen, Larsen, Jones, Andakumar, Banales and Castles2013). The type of reading impairment experienced by poor readers in the NF1 group was categorized using “classification system 5.” Utilizing the CC2, this classification system was evaluated by McArthur, Kohnen, et al. (Reference McArthur, Kohnen, Larsen, Jones, Andakumar, Banales and Castles2013) to be the most appropriate of five different classification systems (including the Edwards and Hogben classification criteria used by Watt et al. (Reference Watt, Shores and North2008) which they applied to the original Castles Word/Nonword test) for poor readers in the general population since it produced similar percentages of each type of reading impairment compared to the other four classification systems; included a clear distinction (half a standard deviation) between lexical and sublexical reading abilities; ensured that a child’s stronger word reading skill was in the average range; and resulted in sufficient group numbers for each type of reading impairment for statistical comparisons (McArthur, Kohnen, et al., Reference McArthur, Kohnen, Larsen, Jones, Andakumar, Banales and Castles2013). The criteria for classification system 5 used to categorize reading impairments in this study were:
-
1. Phonological dyslexia: a z score of or below −1.3 (the 10th percentile or lower) for nonword reading and a z score higher than −1 for irregular word reading.
-
2. Surface dyslexia: a z score of or below −1.3 for irregular word reading and a z score higher than −1 for nonword reading.
-
3. Mixed dyslexia: a z score of −1.3 or lower for both nonword and irregular word readings.
-
4. Unclassified dyslexia: poor readers who were not classified by any of the above criteria. These children were removed from subtype analyses.
Statistical Analysis
Data were analyzed using SPSS version 23 (SPSS Inc., Chicago, IL). Descriptive statistics for continuous normally distributed variables are reported as means (M) and SD or as medians (Mdn) and interquartile ranges for non-normally distributed data. Chi-squared test was used to examine differences between groups for categorical variables. Independent-samples t tests (or Mann–Whitney U tests for non-normally distributed data) were used to compare the NF1 and control groups for cognitive and literacy measures. The risk of being a poor reader in each group (NF1, control) was calculated.
To minimize the number of comparisons made, univariate analyses of variance (ANOVAs) or Kruskal–Wallis (non-normally distributed data) was conducted to identify if there were any differences between all groups (i.e., dyslexic subtypes for NF1, NF1 non-poor readers, and control non-poor readers), with effect sizes calculated (partial eta squared or epsilon squared). For ANOVAs, Welch’s F was used when the homogeneity of variance assumption was violated. For significant group differences, pairwise planned comparisons between (i) the NF1 dyslexic groups and (ii) non-poor readers (NF1 vs. controls) were conducted using independent-samples t test or Mann–Whitney U test. Effect sizes are reported as partial eta squared with values of 0.01, 0.06, and over 0.14 and Cohen’s d values of 0.2–0.4, 0.5–0.7 and ≥0.8, representing small, medium, and large effect sizes, respectively (Cohen, Reference Cohen1988). When scores were non-normally distributed, effect sizes were calculated (r) and converted to Cohen’s d (Fritz, Morris, & Richler, Reference Fritz, Morris and Richler2012). All tests were two-tailed, with level of significance equal to 0.05 and the Holm procedure applied to control the familywise error rate.
In identifying predictors of word reading, to minimize the number of regression models and reduce the number of word reading variables (i.e., nonword, irregular and regular words), a principal component analysis (PCA) was conducted to generate a combined word reading variable. The Kaiser–Meyer–Olkin (KMO) and Bartlett’s test of sphericity was also calculated to check the suitability of the data for PCA. The relationship between combined word reading and cognitive and demographic variables was examined using Pearson correlation coefficient (r), or Spearman rho (r s) for asymmetrically distributed data. Only factors that significantly correlated with combined word reading were entered into the hierarchical multiple regression (stepwise) model according to the strength of their correlation. Because of a significant relationship between IQ and reading ability (Naglieri, Reference Naglieri2001), we controlled for the potential influence of intelligence using an estimate of nonverbal ability (PRI from the WISC-V or PIQ from the WASI-II), which was entered at step 1.
RESULTS
Demographic and cognitive data for both groups are displayed in Table 1. There were no significant between-group differences for sex, age, or SES. Significant group differences were evident for most cognitive measures, with the NF1 group demonstrating significantly poorer performance (with large effect sizes) for IQ measures, visuospatial functioning, and switching attention (all p < .001). Medium effect sizes were observed on measures of divided and sustained attention with performance of the NF1 group falling significantly below controls. There were no significant between-group differences for selective attention. Behaviorally, there was a large effect size for inattentive Attention Deficit Hyperactivity Disorder (ADHD) symptoms with the NF1 group displaying significantly elevated inattentive behaviors.
Table 1. Demographics and means (SD) for general cognitive measures
Conners-3 = Conners’ Rating Scales – Third Edition, Parent Rating Forms; FSIQ = Full-Scale IQ; JLO = Judgement of Line Orientation; PRI/PIQ = Performance IQ; SEIFA = Socio-Economic Indexes for Areas; TEA-Ch = Test of Everyday Attention for Children; VCI/VIQ = Verbal IQ; WASI-II = Wechsler Abbreviated Scale of Intelligence – Second Edition; WISC-IV = Wechsler Intelligence Scale for Children Scale of Intelligence – Fourth Edition, Australian Adaptation; WMI = Working Memory Index.
ªDecile ranking of areas in Australia according to relative socioeconomic advantage and disadvantage; 1 is the lowest and 10 is the highest SES, bmedian, cinterquartile range, dMann–Whitney U test, emissing data for JLO (n = 2) as test not administered, fmissing data for TEA-Ch divided attention subtest (n = 4) as test not administered, gmissing data for TEA-Ch selective attention subtest (n = 2) as test not administered, hmissing data for TEA-Ch switching attention subtest (n = 5) as test not administered, imissing data for Conners-3 (n = 3) as questionnaire not completed.
Holm’s procedure applied to all analyses; higher scores indicate better performance except for Conners-3 where higher scores represent higher levels of ADHD behaviors.
The performance of the NF1 group was significantly poorer than controls on all literacy measures except the LOT, a measure of letter identification (Table 2). Large effect sizes were evident for all other literacy and language variables except for cross-case copying, which yielded a medium effect size.
Within the NF1 group, 49/60 (82%) were classified as poor readers and 11/60 (18%) children as typical readers. In the control group, 7/36 (19%) children were classified as poor readers. The relative risk of poor reading occurring was 4.20 times greater (95% CI 2.14, 8.25) in the NF1 group compared to the control group.
Table 2. Means (SD) for literacy measures for the NF1 and control groups
CC2 = Castles and Coltheart 2; GPC = Grapheme-Phoneme Correspondence; LeST = Letter–Sound Test; LOT = Letter Orientation Test; NEPSY = A Developmental Neuropsychology Assessment; PPVT-4 = Peabody Picture Vocabulary Test – Fourth Edition; TERC = Test of Everyday Reading Comprehension; TOC = Test of Orthographic Choice; TOWRE = Test of Word Reading Efficiency.
ªMedian, binterquartile range, cMann–Whitney U Test.
Holm’s procedure applied to all analyses; higher scores indicate better performance for all measures.
Within the NF1 group of poor readers, 20/49 (41%) met the classification for phonological dyslexia, 24/49 (49%) for mixed dyslexia, and 5/49 (10%) did not meet any classification criteria, despite being poor readers. Literacy measures by subgroup (control non-poor readers, NF1 non-poor readers, NF1 phonological dyslexics, NF1 mixed dyslexics) are displayed in Table 3. While there was no significant difference between the four groups for age, for those variables with raw scores only ANCOVAs were conducted to control for the potential effects of age (Supplementary Table 2). The ANCOVAs resulted in the same pattern of statistical significance for all except one test, the Rain/Hane test, a measure of the phonological lexicon (Supplementary Table 2), which was no longer significantly different between groups after controlling for age. Overall group comparisons indicated significant differences between groups for all literacy variables, except for phonological lexicon, with large effect sizes (Table 3). Planned contrasts indicated the NF1 non-poor readers group performed significantly poorer on measures of nonword reading, nonword repetition, reading comprehension, and semantics when compared to control non-poor readers. Planned contrasts between the NF1 dyslexic groups indicated that children with mixed dyslexia performed significantly poorer on measures of regular word, irregular word, and nonword reading, sight word and phonemic decoding efficiency, letter–sound knowledge, parsing, and reading comprehension when compared to children with phonological dyslexia (Table 3).
Table 3. Means (SD) literacy scores by reading subtypes
ηp2 = partial eta-squared; CC2, Castles and Coltheart 2; GPC, Grapheme-Phoneme Correspondence; LEST, Letter–Sound Test; NEPSY, A Developmental Neuropsychology Assessment; PPVT-4, Peabody Picture Vocabulary Test – Fourth Edition; TERC, Test of Everyday Reading Comprehension; TOC, Test of Orthographic Choice; TOWRE, Test of Word Reading Efficiency; SS, standard score.
*p ≤ .05, **p ≤ .01.
aChi-square test of independence, bmedian, cinterquartile range, dKruskal–Wallis H Test, eCohen’s d, fWelch’s F, gMann–Whitney U Test, hmissing data (n = 1) as test not administered, iANCOVA results controlling for age (see Supplementary Table 2).
The KMO value (KMO = 0.73) indicated that the sampling adequacy of the data was suitable for PCA. Bartlett’s test of sphericity also indicated that between-items correlations were appropriate for PCA (χ 2(3) = 299.70, p <. 001). Analysis of the eigenvalues indicated three components with only one having an eigenvalue over Kaiser’s criterion of 1, explaining 90.45% of the variance. Hence, this was the only component retained. This new variable, combined word reading, was examined for normality and univariate outliers using a cutoff of three SDs outside the mean. Two participants had scores that fell outside this range. To reduce the influence of these outliers, the combined word reading variable was transformed (Field, Reference Field2009). However, after transformation, the outliers still remained. Hence, as recommended by Field (Reference Field2009, p. 153), the scores for these two participants were changed to one unit above the next highest score for combined word reading variable. There was no significant difference between males (M = −0.51, SD = 0.66) and females (M = −0.50, SD = 0.62) on the combined word reading variable (t (58) = −0.07, p = .95, d = 0.02) or any of the individual measures of reading subskills (all p>.07; Supplementary Table 1).
Stronger combined word reading was moderately associated with better receptive language, working memory, and higher FSIQ and verbal IQ (Table 4). There was a significant moderate negative relationship between combined word reading and inattentive behaviors indicating that children with elevated inattentive ADHD symptoms had poorer combined word reading. There was no significant relationship between combined word reading and SES or visuospatial functioning.
Table 4. Correlations between word reading and cognitive, demographic, and behavioral variables for the NF1 group
FSIQ = Full-scale IQ; JLO = Judgement of Line Orientation; PPVT-4 = Peabody Picture Vocabulary Test – Fourth Edition; PRI/PIQ = Performance IQ; SEIFA = Socio-Economic Indexes for Areas; VCI/VIQ = Verbal IQ; WMI = Working Memory Index.
*p < .05, **p < .01.
aSpearman’s rho, bdecile ranking of areas in Australia according to relative socioeconomic advantage and disadvantage; 1 is the lowest and 10 is the highest SES, cn = 58, dn = 57.
In the final regression model, working memory, inattentive behaviors, and receptive language were significant predictors of combined word reading, accounting for 31% of the variance (Table 5).
Table 5. Hierarchical regression analysis predicting word reading in the NF1 group
Performance IQ = PRI/PIQ; B = unstandardized beta; SE B = standard error of B; β = standardized beta.
**p ≤ .01, *p ≤ .05.
DISCUSSION
Consistent with our prediction, children with NF1 demonstrated poorer performance than controls on all literacy measures except letter orientation. Examination of the NF1 group revealed that 82% displayed poor word reading compared to 19% of controls. The relative risk of being a poor word reader was 4.20 times greater for children with NF1 compared to controls, providing support for previous findings that children with NF1 are at high risk of reading difficulties (Cutting et al., Reference Cutting, Koth and Denckla2000; Watt et al., Reference Watt, Shores and North2008).
This is the first time the DRC model has been used as a framework to examine the cognitive processes underlying word reading in children with NF1. Results indicated that nonword reading was the most severely impaired literacy skill in the NF1 group. The NF1 group also demonstrated significant weaknesses in irregular word reading and lexical subskills (i.e., orthographic lexicon, phonological lexicon, semantics). Poor performance was also identified for subskills common to both reading routes (i.e., letter identification, blending, nonword repetition). These findings indicate that children with NF1 present with widespread difficulties across subskills that map onto lexical and sublexical reading routes.
Within the NF1 poor reading group, there was a fairly comparable breakdown of children with phonological (41%) and mixed dyslexia (49%). Poor readers with NF1 were more likely to have phonological dyslexia than that observed in poor readers from the general population (McArthur, Kohnen, et al., Reference McArthur, Kohnen, Larsen, Jones, Andakumar, Banales and Castles2013). Similar to Watt et al. (Reference Watt, Shores and North2008), we found no NF1 children with surface dyslexia. Poor readers with NF1 displayed a similar rate of mixed dyslexia to that observed in poor readers in the general population (McArthur, Kohnen, et al., Reference McArthur, Kohnen, Larsen, Jones, Andakumar, Banales and Castles2013). Children with phonological dyslexia performed similarly to those with mixed dyslexia although the latter demonstrated more widespread difficulties in reading subskills and, overall, performed more poorly. Critically, children with mixed dyslexia displayed significant weaknesses in reading fluency and comprehension, indicating a greater functional impairment than children with phonological dyslexia.
Although we found an unusually high percentage of phonological dyslexics in our NF1 sample, the proportion was lower than anticipated and did not support our hypothesis that the majority of poor readers would present with phonological dyslexia. While it is not clear why our findings differed from a previous study (Watt et al., Reference Watt, Shores and North2008), potential contributing factors include utilizing a revised version of the CC2 test and applying a different classification system for categorizing dyslexia subtypes. Here, rather than the Edwards and Hogben (Reference Edwards and Hogben1999) criteria used by Reference Watt, Shores and NorthWatt et al. (2008), we used a modified criteria (classification 5) reported by McArthur, Kohnen, et al. (Reference McArthur, Kohnen, Larsen, Jones, Andakumar, Banales and Castles2013). Interestingly, when McArthur, Kohnen, et al. (Reference McArthur, Kohnen, Larsen, Jones, Andakumar, Banales and Castles2013) compared classification system 5 to the Edwards and Hogben criteria using CC2 from the same group of children, they found classification 5 returned a threefold increase in the percentage of poor readers identified as having phonological dyslexia compared to the Edwards and Hogben criteria. Therefore, it seems unlikely that the use of a different classification system contributed to the lower percentage of phonological dyslexia we observed compared to Watt et al. (Reference Watt, Shores and North2008). It is also possible that our findings are more representative of the broader NF1 population as the sample size for the current study was twice the size of Watt et al. (Reference Watt, Shores and North2008).
The reason why children with NF1 experience such an unusually high incidence of phonological dyslexia is unclear. Evidence from the general population indicates a strong genetic basis to phonological dyslexia and that it may be related to an underlying impairment in processing spoken language (Castles, Datta, Gayan, & Olson, Reference Castles, Datta, Gayan and Olson1999). It has been previously suggested that a phonological impairment may be an inherent feature of NF1 (Chaix et al., Reference Chaix, Lauwers-Cancès, Faure-Marie, Gentil, Lelong, Schweitzer and Castelnau2017) which is supported by previous findings indicating that children with NF1 (5–6 years old) were 5.60 times more likely to display phonological deficits than unaffected children (Arnold et al., Reference Arnold, Payne, Lorenzo, North and Barton2018). Abnormalities in brain structure (Payne, Moharir, Webster, & North, Reference Payne, Moharir, Webster and North2010) and abnormal cell signaling (Costa et al., Reference Costa, Federov, Kogan, Murphy, Stern, Ohno and Silva2002) in NF1 may contribute to phonological impairments. While there are no conclusive links between neurobiological abnormalities in NF1 and poor reading, one study reported greater symmetry in the planum temporale of boys with NF1 and that lack of typical asymmetry was related to poorer reading performance (Billingsley, Schrimsher, Jackson, Slopis, & Moore, Reference Billingsley, Schrimsher, Jackson, Slopis and Moore2002). However, it should be noted that the participants in the Billingsley et al. (Reference Billingsley, Schrimsher, Jackson, Slopis and Moore2002) study were not reading impaired and presented with intact phonological processing. Further research is needed to clarify possible relationships between the neurobiological and cognitive abnormalities reported in NF1.
As expected, verbal working memory skills were a significant predictor of reading ability of children with NF1. Impairments in working memory are a common feature in children with NF1 (Payne, Arnold, Pride, & North, Reference Payne, Arnold, Pride and North2012). Our findings suggest that children with reduced working memory capacity are at an increased risk of reading difficulties, which is consistent with findings in the general population (Wang & Gathercole, Reference Wang and Gathercole2013). Stronger working memory abilities support complex, higher level processing (Pham & Hasson, Reference Pham and Hasson2014) and are proposed to play an important role when several cognitive processes are concurrently involved (e.g., when decoding words, accessing semantic knowledge and retrieving information previously read) (Sesma, Mahone, Levine, Eason, & Cutting, Reference Sesma, Mahone, Levine, Eason and Cutting2009). As a consequence, greater working memory capacity has been linked to better reading fluency and comprehension as well as basic reading abilities (Pham & Hasson, Reference Pham and Hasson2014; Sesma et al., Reference Sesma, Mahone, Levine, Eason and Cutting2009). There is also evidence to suggest that children with NF1 who have stronger working memory capacity experience greater benefits from reading intervention than those with poorer working memory (Arnold, Barton, McArthur, North, & Payne, Reference Arnold, Barton, McArthur, North and Payne2016).
Comorbidity between ADHD symptoms and literacy problems has previously been reported in the general population (Peterson & Pennington, Reference Peterson and Pennington2015) and in children with NF1 (Hyman, Shores, & North, Reference Hyman, Shores and North2006; Pride, Payne, & North, Reference Pride, Payne and North2012). Consistent with this, we found that elevated inattentive symptoms were significantly associated with poorer word reading in children with NF1. It may be beneficial to treat comorbid ADHD symptoms in conjunction with targeted literacy intervention. However, in the general population, there is a mixed support for the transfer effects of stimulant medication upon children’s literacy abilities (Kortekaas-Rijlaarsdam, Luman, Sonuga-Barke, & Oosterlaan, Reference Kortekaas-Rijlaarsdam, Luman, Sonuga-Barke and Oosterlaan2018). While there is some evidence that methylphenidate can improve ADHD symptoms in children with NF1 (Lion-François et al., Reference Lion-François, Gueyffier, Mercier, Gérard, Herbillon, Kemlin and Kassaï2014; Mautner, Kluwe, Thakker, & Leark, Reference Mautner, Kluwe, Thakker and Leark2002), no studies have examined the effects of stimulant medication on reading outcomes in NF1.
We found that the word reading skills of children with NF1 were not significantly related to sex or SES. We found a weak association between SES and academic achievement (r s = 0.21 converted r = 0.22 (Gilpin, Reference Gilpin1993)), which is consistent with another Australian study that also used the SEIFA (Marks, Reference Marks2017) and two meta-analyses indicating an SES–achievement correlation value ranging from 0.22 (White, Reference White1982) to 0.25 (Harwell, Maeda, Bishop, & Xie, Reference Harwell, Maeda, Bishop and Xie2017). Our measure of SES (SEIFA) is area based (i.e., geographical location of the child’s family home), which may potentially underestimate the SES–achievement correlation. An Australian study found that correlations of SEIFA indexes with achievement were approximately one-third less than the correlations between composite individual levels of SES and achievement (Ainley & Long, Reference Ainley, Long, Ainley, Graetz, Long and Batten1995). It is possible that a measure of SES that assessed individual aspects of the family such as parental years of education, income, and/or occupation may have produced different results (Diemer, Mistry, Wadsworth, López, & Reimers, Reference Diemer, Mistry, Wadsworth, López and Reimers2013; Hauser, Reference Hauser1994; Marks, Reference Marks2017). Also our SES findings might not generalize to areas outside of Australia, especially in other parts of the world that have greater wealth disparities. In addition, we found no evidence of a relationship between visuospatial skills (JLO) in children with NF1 and reading. The relationship observed by Cutting and Levine (Reference Cutting and Levine2010) was for only a small group of children with NF1 and reading disabilities (n = 12). While results from a study suggest that the visual magnocellular pathway is impaired in individuals with NF1 (Violante et al., Reference Violante, Ribeiro, Cunha, Bernardino, Duarte, Ramos and Castelo-Branco2012), their reading abilities were not assessed. Further research is needed to clarify the relationship between reading and visuospatial abilities of children with NF1.
This study is not without limitations. We acknowledge that there are a number of other models of reading (i.e., Connectionist; Harm & Seidenberg, Reference Harm and Seidenberg1999; Seidenberg & McClelland, Reference Seidenberg and McClelland1989). However, the DRC was utilized in this study as it is a particularly useful theoretical basis for conceptualizing links between particular reading impairments and the subskills that underlie them. Furthermore, it has been shown to be appropriate for detecting the high occurrence of phonological deficits in children with NF1 (Watt et al., Reference Watt, Shores and North2008). Consequently, we feel the results of this study have important contributions to make regarding the nature of reading difficulties in school-age children with NF1.
The study findings highlight the high incidence of phonological decoding difficulties occurring in children with NF1. Previous research from the general population (Galuschka, Ise, Krick, & Schulte-Korne, Reference Galuschka, Ise, Krick and Schulte-Korne2014) and in children with NF1 (Arnold et al., Reference Arnold, Barton, McArthur, North and Payne2016) indicates that most poor readers will benefit from phonics training which explicitly teaches letter–sound knowledge. Our results highlight the importance of a thorough assessment of the reading abilities and underlying subskills in children with NF1 so that children can receive appropriate, targeted intervention. Finally, the relationship identified between reading and other cognitive abilities (i.e., working memory, language, attention) emphasizes the need for a combined assessment of literacy and broader cognitive skills.
Further investigation of the reading subskills of children with NF1 who are poor readers is needed to better understand the differences between children with predominately phonological impairments and those with mixed dyslexia. There is some evidence from the general population indicating that children with mixed dyslexia may be the most severely impaired (McArthur, Kohnen, et al., Reference McArthur, Kohnen, Larsen, Jones, Andakumar, Banales and Castles2013). Case studies of children with mixed dyslexia in the general population indicate that although there may be some generalization of specific training of sublexical skills to lexical skills (Brunsdon, Hannan, Nickels, & Coltheart, Reference Brunsdon, Hannan, Nickels and Coltheart2002), there is also independence between the two reading routes (Brunsdon et al., Reference Brunsdon, Hannan, Nickels and Coltheart2002). Hence, children with mixed dyslexia in our NF1 sample may require more intensive intervention than those with phonological dyslexia.
In conclusion, we confirm an elevated risk of reading impairment in children with NF1; over 80% of children with NF1 in our sample displayed poor word reading and they exhibited a 4.20 times greater risk of experiencing a reading difficulty than unaffected controls. For the first time, we have shown that deficits occur throughout lexical and sublexical reading subskills and also in basic reading precursor skills, such as letter identification. Higher level literacy processes such as reading fluency and comprehension are also commonly impaired. The high frequency of word reading impairments and study findings from others (Cutting et al., Reference Cutting, Clements, Lightman, Yerby-Hammack and Denckla2004; Mazzocco et al., Reference Mazzocco, Turner, Denckla, Hofman, Scanlon and Vellutino1995) indicate that the cognitive phenotype of NF1 cannot be adequately described as being consistent with a nonverbal learning disability profile as previously proposed (Denckla, Reference Denckla1996). Poor readers with NF1 appear to be at increased risk for phonological dyslexia compared to peers in the general population (McArthur, Kohnen, et al., Reference McArthur, Kohnen, Larsen, Jones, Andakumar, Banales and Castles2013).
ACKNOWLEDGMENTS
The authors received no financial support for this research.
CONFLICT OF INTEREST
The author(s) have nothing to disclose.
SUPPLEMENTARY MATERIAL
To view supplementary material for this article, please visit https://doi.org/10.1017/S135561772000106X
