Introduction to the Special Issue “Comorbidities between Reading Disorders and Other Developmental Disorders”

Reading disorder frequently co-occurs with other neurodevelopmental disorders and this has important implications for theory as well as practice. Understanding the nature and causes of comorbidities is at the heart of understanding developmental disorders. For example, if Disorder X frequently co-occurs with Disorder Y, this points to the operation of shared risk factors and improves understanding of the causes of each of these disorders. This special issue addresses recent advances in understanding comorbidities between reading disorder/dyslexia and other developmental disorders and identifies issues requiring further research.

Comorbidity can be found between disorders within the same diagnostic grouping (homotypic comorbidity: e.g., reading disorder and mathematics disorder), as well as between disorders from different diagnostic groupings (heterotypic comorbidity), such as between reading disorder and behavioral disorders (Attention-Deficit-Hyperactivity-Disorder (ADHD) and conduct disorder) or reading disorder and emotional problems (anxiety and depression) (Angold, Costello, & Erkanli, 1999). Rates of comorbidity between reading disorder and other neurodevelopmental disorders range between 11% and 70% for reading disorder and mathematics disorder (Moll, Kunze, Neuhoff, Bruder, & Schulte-Körne, 2014, for an overview), between 20% and 50% for reading disorder and behavioral disorders (ADHD and conduct disorder) (Hendren, Haft, Black, White, & Hoeft, 2018; Hinshaw, 1992; Margari et al., 2013; Sexton, Gelhorn, Bell, & Classi, 2012; Willcutt & Pennington, 2000), and between 9% and 29% for reading disorder and emotional problems (i.e. anxiety disorder) (Carroll, Maughan, Goodman, & Meltzer, 2005; Margari et al., 2013; Maughan, Rowe, Loeber, & Stouthamer-Loeber, 2003). Although comorbidity rates vary widely between studies, they are clearly consistently higher than expected by chance, indicating that comorbidities between reading disorder and other neurodevelopmental disorders are the rule rather than an exception. However, it is important to note that reading disorder also occurs in isolation, as do other neurodevelopmental disorders. This raises questions about how the risk factors underlying reading and other neurodevelopmental disorders interact and how to explain the different behavioral patterns. A comprehensive model of reading disorder (or any other developmental disorder) needs to be able to explain not only the disorder itself but also its co-occurrence with other disorders (see Landerl, Göbel, & Moll, 2013 for a similar discussion on mathematics disorder). Nevertheless, most research over the last few decades has focused on single disorders, neglecting co-occurring features. In fact, some studies have deliberately excluded individuals with additional problems (notably, problems in attention) in order to identify the deficits that are specific to a given disorder.

A major step toward a better understanding of developmental disorders and their comorbidities was the shift from single deficit models to multiple deficit models, first discussed in the ground-breaking work by Bruce Pennington (2006). While earlier research based on single-deficit models had helped to identify the deficits that are specific to a given disorder (such as reading disorder), it became clear that single-deficit approaches suffer from several shortcomings and a new theoretical framework was required to account for the complex etiology of neurodevelopmental disorders (Bishop, 2008; Pennington, 2006; Rutter, 2006; Snowling, 2012). First, there is increasing evidence that the etiology of such disorders is multifactorial, and cannot be explained by single causes at the genetic, environmental, neurobiological or cognitive levels of analysis. Second, single-deficit models do not explain the heterogeneity in symptomatology observed between individuals diagnosed with the same disorder. Finally, and most importantly for the current special issue, single-deficit models do not account for the high comorbidity rates observed between different neurodevelopmental disorders.

Single-deficit models are especially problematic for explaining the high comorbidity rates between disorders from different diagnostic groupings, such as the comorbidity between reading disorder and ADHD (heterotypic comorbidity); this is because the two disorders are characterized by clearly distinct core deficits (phonological deficits in the case of reading disorder and executive deficits in ADHD). Within a single-deficit framework, one way to explain the comorbidity between reading disorder and ADHD is the symptom phenocopy hypothesis (Neale & Kendler, 1995): this suggests that reading disorder is the primary disorder and the symptoms of ADHD are secondary symptoms (themselves caused by poor reading) or that ADHD is the primary disorder causing symptoms of poor reading. Another potential explanation is the three-independent-disorders hypothesis, suggesting that the comorbid disorder (e.g., RD+ADHD) represents a third disorder that is etiologically distinct from the single disorders (reading disorder only and ADHD only). However, neither of these hypotheses has been well supported in the case of reading disorder and ADHD (e.g., Adams & Snowling, 2001; Pennington, Groisser, & Welsh, 1993; Willcutt et al., 2001). Similarly, previous research has shown neither explanation can account for homotypic comorbidity (i.e. reading disorder and mathematics disorder) (Moll, Göbel, & Snowling, 2015; Raddatz, Kuhn, Holling, Moll, & Dobel, 2016).

Together, these shortcomings have led to the development of multiple-deficit models, which suggest that neurodevelopmental disorders result from interactions between multiple biological and environmental risk and protective factors which together determine the behavioral outcome of a putative disorder (Pennington, 2006; Pennington et al., 2012). Within the multiple-deficit framework, comorbidities are explained by shared risk factors that have effects on more than one outcome. For example, a cognitive deficit in language skills or an environmental risk factor such as low socioeconomic status can affect learning across several domains, resulting in problems in both, reading and math. In addition, comorbidity can also result from overlapping symptoms, meaning that the same behavioral outcome can be affected by different cognitive risk factors. For example, poor performance on mathematical word problems or fact retrieval (i.e. multiplication facts) can be caused by poor number processing skills, or a consequence of poor language or reading skills (De Smedt & Boets, 2010; Donlan, Cowan, Newton, & Lloyd, 2007; Goebel & Snowling, 2010; Koponen, Mononen, Rasanen, & Ahonen, 2006; Moll et al., 2015; Moll, Landerl, Snowling, & Schulte-Körne, 2019; Simmons & Singleton, 2006).

Multiple-deficit approaches also account for the fact that the core deficits associated with a specific neurodevelopmental disorder are not deterministic. For example, although a deficit in phoneme awareness represents a cognitive core deficit underlying reading disorder, such a deficit is neither necessary nor sufficient to explain the disorder. Thus, not all individuals with reading disorder show a phoneme awareness deficit (e.g., Moll & Landerl, 2009; Pennington et al., 2012; Wimmer & Mayringer, 2002), and not all individuals with a phoneme awareness deficit have reading disorder (e.g., Moll, Loff, & Snowling, 2013; Moll et al., 2016; Snowling, 2008; Snowling, Gallagher, & Frith, 2003; van Bergen, de Jong, Plakas, Maassen, & van der Leij, 2012). Furthermore, some risk factors, especially those representing a core deficit, will be more relevant than others in relation to a specific symptom (e.g. poor nonword reading) and hence need to be weighted more heavily in a model determining a diagnostic threshold (e.g. for reading disorder). In short, probabilistic risk factors accumulate to increase the likelihood of fulfilling diagnostic criteria for a specific disorder (Hulme & Snowling, 2013). These risk factors can be identified on different levels (genetic, environmental, brain and cognitive levels) which interact, resulting in complex many-to-many mappings between the different levels of analysis (Kendler, 2012). The commentary by Mc Grath and Pennington (this volume) discusses multiple deficit framework models and highlights the progress that has been made in understanding comorbidity based on the multiple deficit framework, as well as constraints and perspectives for further research on comorbidity.

Another important feature of many of the risk and protective factors associated with development are that they are not “all-or-none” but rather are continuous. In line with this idea, reading disorder and other developmental disorders are best conceptualized in dimensional terms each representing the lower end of the normal distribution for the relevant skill. Therefore, cutoff criteria used to define reading or other developmental disorder are arbitrary and typically defined according to externally agreed criteria (Branum-Martin, Fletcher, & Stuebing, 2013). In practice, as Joyner and Wagner (this volume) argue, classification around the cut-point can be particularly unstable. One way to address this issue is to take advantage of co-occurring difficulties in order to increase diagnostic reliability. Joyner and Wagner followed this idea and used a meta-analytic approach to argue for the utility of including math disability as an additional predictor for reading disorder.

Another way around the problem of categorical diagnosis is to investigate the correlation between normally distributed skill dimensions, such as reading and mathematics. The paper by Daucourt, Erbeli, Haughbrook, Little, and Hart (this volume) follows this approach to investigate the etiological influences on the overlap between reading, mathematics and attention disorders. This paper reports a meta-analysis of 38 twin studies of children from third to ninth grade (children with normally varying traits and subgroups with cutoffs) who were assessed for reading, mathematics and attention skills. Across these studies, 60% of children with reading disorder had either mathematics disorder or ADHD, the phenotypic correlations between reading and mathematics being stronger than that between reading and ADHD symptoms. A key question was the amount of variance in the comorbidity between disorders, which was attributable to genetic and environmental factors. For reading and mathematics, there was significant shared genetic variance accounting for comorbidity as well as substantial variance attributable to shared and non-shared environmental factors. The estimates were lower for the comorbidity between reading and ADHD. In both cases, however, there was considerable heterogeneity between studies due to a range of factors including the specific assessments and response formats used. While the findings require replication, they converge with the view that reading and mathematics disorders share common risk factors while reading and ADHD are distinct disorders.

A limitation of much of the evidence on the comorbidities of reading disorder is that it is based on data from cross-sectional studies. If we are to understand whether comorbid disorders develop in parallel with, or as secondary consequences of co-occurring disorders (as implied by the phenocopy hypothesis), it is important to examine longitudinal data. The paper by Horbach, Scharke, Heim, and Günther (this volume) provides a developmental perspective on the relationship between specific learning disorder and co-occurring behavioral and emotional difficulties. In this longitudinal study from Kindergarten to fifth grade, the authors assessed the development of emotional and behavioral problems in children diagnosed with reading and spelling disorder using parental ratings of children’s behavior on the Child Behavior Checklist (CBCL) (Döpfner et., 1988). The authors modeled these data to investigate changes in behavior over time using growth component modeling. Although there were no significant differences between children with specific learning disorder (SLD) and those without SLD in preschool, there was an increase in emotional and behavioral difficulties after school entrance, such that the group difficulties became significant and remained so over time, with a slight drop after secondary school entrance. However, about a third of the sample in this study fulfilled the criterion for ADHD. When the data were examined separately for those with and without attention problems, the group differences between children with and without SLD were no longer significant. Arguably, these findings are somewhat tautologous, given that parental ratings at least in part reflect attention problems. However, there was a similar pattern for internalizing problems (e.g., anxiety) as for the externalizing behaviors more typically associated with ADHD. Notwithstanding the small sample of children with ADHD and the need for parental ratings to be validated (by triangulating with teacher judgments), these findings underline the importance of tracking children through time and particularly through vulnerable phases (e.g., transition from preschool to primary, and primary to secondary school) if the impact of comorbidities is to be understood.

One of the most important reasons for understanding the causes of neurodevelopmental disorders is to devise theoretically motivated interventions. More specifically, there is a significant question to be answered regarding whether reading interventions, which are known to be effective for children with pure disorders, are as effective for those with comorbid disorders (Hendren et al., 2018; Sexton et al., 2012). Related questions include Should individuals with comorbid disorders receive interventions for both disorders and is it best to provide the two different treatments in parallel or sequentially?, and Are there any transfer effects of a specific intervention to the symptomatology of a comorbid disorder? Denton, Tamm, Schatschneider, and Epstein (this volume) address this latter issue by asking whether for children with comorbid reading disorder and ADHD, an intervention to treat attentional problems leads to improvements in reading skills and whether combining reading intervention with treatment for ADHD has a synergistic effect on reading outcomes (making the intervention more effective). To investigate these issues, the authors examined the effects of three different interventions (i.e. ADHD treatment, reading intervention and combined ADHD and reading treatment) on the reading skills of children with comorbid ADHD and reading disorder in a randomized controlled trial. The treatment for attention and the intervention for reading had largely independent effects on attention and on aspects of reading fluency. On reading comprehension, the effects were however interestingly different in that the attention treatment had a more positive effective than reading intervention. Finally, in neither case, was the combined treatment better than either treatment alone. The findings, although preliminary, warrant replication in order to better understand the mechanisms that account for the gains in reading comprehension following on from ADHD treatment.

The articles and commentary presented here illustrate the great progress that has been made in recent years in understanding comorbidity. However, the findings also highlight a number of open questions and approaches for future research. These include (1) the integration of different levels (genetic, environmental, brain and cognitive levels) when analyzing comorbidity between reading disorder and co-occurring difficulties, (2) the analysis of developmental changes in the association of risk (and protective) factors underlying reading disorder and comorbid disorders, and (3) the improvement of diagnostic processes and of treatments for individuals with two or more comorbid disorders.