Neuropsychological analyses of RD and ADHD 1 RUNNING HEAD: NEUROPSYCHOLOGY OF RD AND ADHD In Press, Developmental Neuropsychology Neuropsychological analyses of comorbidity between reading disability and attention-deficit/hyperactivity disorder: In search of the common deficit

Measures of component reading and language skills, executive functions, and processing speed were administered to groups of children with attention-deficit hyperactivity disorder (ADHD; n = 113), reading disability (RD; n = 109), both RD and ADHD (n = 64), and neither RD nor ADHD (n = 151). Groups with RD exhibited pronounced deficits on all measures of component reading and language skills, as well as significant weaknesses on measures of verbal working memory, processing speed, and response inhibition. Groups with ADHD exhibited weaknesses on all response inhibition and processing speed tasks and were impaired on some measures of component reading skills and verbal working memory. The group with comorbid RD and ADHD exhibited the combination of the deficits in the RD-only and ADHD-only groups, providing evidence against the phenocopy and cognitive subtype hypotheses as explanations for the cooccurrence of RD and ADHD. Slow and variable processing speed was characteristic of all three clinical groups, suggesting that measures of this domain may be useful for future studies that search for the common genes that increase susceptibility to RD and ADHD. Neuropsychological analyses of RD and ADHD 3 Neuropsychological analyses of comorbidity between RD and ADHD: In search of the common deficit Reading disorder (RD) and attention-deficit/hyperactivity disorder (ADHD) are two of the most common disorders of childhood, each occurring in approximately five percent of the population (American Psychiatric Association, 2000). ADHD and RD also co-occur significantly more frequently than expected by chance; 25 40% of individuals with ADHD also meet criteria for RD (e.g., Dykman & Ackerman, 1991; Semrud-Clikeman et al., 1992), whereas 15 40% of individuals with RD meet criteria for ADHD (Gilger, Pennington, & DeFries, 1992; Shaywitz, Fletcher, & Shaywitz, 1995; Willcutt & Pennington, 2000). Several competing explanations have been proposed to account for comorbidity of RD and ADHD. We first briefly review data that do not support several of these hypotheses, then describe four additional hypotheses that have received empirical support. We then summarize the implications of neuropsychological studies for these competing hypotheses, and provide an overview of current knowledge regarding the neuropsychological correlates of RD, ADHD, and comorbid RD+ADHD. Artifactual explanations of comorbidity between RD and ADHD Before attempting to understand the etiological underpinnings of comorbidity between disorders, it is important to rule out the possibility that the observed comorbidity is an artifact that is caused by a biased sampling procedure or measurement problem. For example, artifactual comorbidity could occur due to ascertainment biases in clinic-referred samples, common method variance in the measures used to define the disorders, symptom overlap between the disorders, or rater bias. Most of these artifactual hypotheses can be rejected based on existing data. RD and ADHD co-occur more frequently than expected by chance in both samples ascertained from clinics (e.g., Semrud-Clikeman et al., 1992) and non-referred samples recruited from the community (e.g., Fergusson & Horwood, 1992; Willcutt & Pennington, 2000), indicating that this comorbidity is not restricted to clinic-referred samples. Because RD is assessed by cognitive tests whereas ADHD is assessed by behavioral ratings, the relation between RD and ADHD cannot be explained by shared method variance. Similarly, the symptoms of RD and ADHD as defined in the fourth edition of the Diagnostic and Statistical Manual of Mental Disorders (DSM-IV; American Psychiatric Association, 1994, 2000) do not overlap. The rater-bias hypothesis is somewhat more difficult to test, and the possibility remains that parents or teachers may be more likely to endorse ADHD symptoms if they know that the child is experiencing difficulty with reading. However, previous studies indicate that in addition to higher ratings of inattention symptoms by parents and teachers, children with RD report greater difficulties with attention than children without RD on self-report measures (Willcutt, Chhabildas, & Pennington, 1998). Although the rater-bias hypothesis cannot be conclusively rejected based on these results, these data suggest that it is not likely to provide a sufficient explanation for all cases of comorbidity between RD and ADHD. Competing explanations for true comorbidity Numerous competing hypotheses have been proposed to explain non-artifactual comorbidity between two disorders (e.g., Caron & Rutter, 1991; Neale & Kendler, 1995). Explanations that have been supported in some previous studies of RD and ADHD include the cross-assortment hypothesis (e.g., Faraone et al., 1993), the phenocopy hypothesis (e.g., Pennington, Groisser, & Welsh, 1993), the cognitive subtype hypothesis (e.g., Rucklidge & Tannock, 2002), and the common etiology hypothesis (e.g., Willcutt et al., 2003). The cross-assortment hypothesis suggests that an individual with RD is more likely to have a child with an individual with ADHD than would be expected by chance based on population base rates of RD and ADHD. In a family study of the biological relatives of children with ADHD, Faraone et al. (1993) found that comorbidity between learning disabilities and ADHD was best explained by cross-assortment. This result, however, was not replicated in later studies (Doyle, Faraone, DuPre, & Biederman, 2001; Friedman, Chhabildas, Budhiraja, Willcutt, & Pennington, 2003), suggesting that cross-assortment is not likely to explain the majority of cases of comorbid RD and ADHD. Pennington et al. (1993) described results in a small sample of children with RD and ADHD that suggested that RD might lead to the phenotypic manifestation of ADHD in the absence of the etiological influences typically associated with ADHD in isolation. This phenocopy hypothesis suggests that a child might appear to be inattentive or hyperactive in the classroom due to frustration elicited by difficulties with reading, rather than as a consequence of the neurocognitive difficulties that are typically associated with ADHD in the absence of RD. More recent data from larger samples, however, have generally failed to support the Neuropsychological analyses of RD and ADHD 4 phenocopy hypothesis (e.g., Nigg, Hinshaw, Carte, & Treuting, 1998; Rucklidge & Tannock, 2002; Seidman, Biederman, Monuteaux, Doyle, & Faraone, 2001; Willcutt, Pennington, Boada, et al., 2001). The cognitive subtype hypothesis suggests that comorbid RD+ADHD is a third disorder that is due at least in part to etiological factors that are distinct from those that increase susceptibility to RD or ADHD alone. Therefore, this hypothesis predicts that the comorbid group will exhibit a different pattern of external correlates than would be expected based on the additive combination of the correlates of each disorder when they occur separately. Rucklidge and Tannock (2002) found that the comorbid group performed significantly worse than the RD-only and ADHD-only groups on measures of color naming, providing some support for this hypothesis. In contrast, other studies found that the RD+ADHD group exhibited the additive combination of the deficits associated with each individual disorder (e.g., Pisecco, Baker, Silva, & Brooke, 2001; Swanson, Mink, & Bocian, 1999; Willcutt et al., 2001), suggesting that additional research is needed. Finally, a series of studies tested if the relation between RD and ADHD is attributable to common etiological influences that increase susceptibility to both disorders. Twin studies indicate that RD and ADHD are each highly heritable and polygenic (e.g., DeFries & Alarcón, 1996; Gayán & Olson, 2001; Faraone, Doyle, Mick, & Biederman, 2001; Fisher & DeFries, 2002; Levy, Hay, McStephen, Wood, & Waldman, 1997; Willcutt, Pennington, & DeFries, 2000a; Willcutt, in press), and bivariate twin analyses suggest that comorbidity between RD and ADHD is due primarily to common genetic influences (Light, Pennington, Gilger, & DeFries, 1995; Stevenson, Pennington, Gilger, DeFries, & Gillis, 1993; Willcutt, Pennington, & DeFries, 2000b; Willcutt, DeFries, et al., 2003). Based on these results, genetic linkage studies have begun to search for chromosomal regions that may contain a gene or genes that increases risk for both disorders (Loo et al., in press; Willcutt et al., 2002, 2003). In the first of these studies, Willcutt et al. (2002, 2003) reported that the well-replicated quantitative trait locus for RD on chromosome 6p21 (e.g., Cardon et al., 1994, 1995) also increases susceptibility to ADHD. Similarly, Loo et al. (in press) screened the entire genome and found that regions of chromosomes 16p and 17q may contain genes that increase risk for both RD and ADHD. Thus, although several of these results await independent replication, existing data provide the strongest support for the hypothesis that comorbidity between RD and ADHD is due at least in part to a common genetic etiology. Although these previous studies suggest that common genetic influences contribute to comorbidity of RD and ADHD, the mechanisms of these common genes are unknown. Because a single gene can influence multiple facets of neurocognitive development (e.g., Falconer & MacKay, 1996), it is possible that the same genetic influences could increase risk for RD due to their effects on a first neural system and increase risk for ADHD due to the impact of the gene on a second, distinct pathophysiological substrate. In this case there would be no common neuropsychological deficit in groups with RD and ADHD. Alternatively, a more parsimonious model would suggest that the common genes lead to a developmental change in a single pathophysiological substrate, and that this change then increases risk for both disorders. In this model all three groups (i.e., RD, ADHD, and RD+ADHD) should be characterized by at least some common neuropsychological weaknesses, with each individual’s final phenotype determined by the other genetic and environmental influences that affect that individual. ADHD symptom dimensions. A final factor that complicates interpretation of previous studies of comorbidity of RD and ADHD is the distinction between symptoms of inattention and hyperactivity-impulsivity in DSM-IV. Phenotypic analyses suggest that deficits in reading achievement and more general academic difficulties are more strongly associated with inattention than hyperactivity-impulsivity (e.g., Chhabildas, Pennington, & Willcutt, 2001; Lahey & Willcutt, 2002; Molina, Smith & Pelham, 2001; Willcutt & Pennington, 2000; Wolraich, Feurer, Hannah, Baumgaertel, & Pinnock, 1998). Similarly, twin studies suggest that whereas the phenotypic correlation between reading deficits and inattention symptoms is primarily explained by common genetic influences, these common genes play a smaller role in the correlation between reading deficits and hyperactivity-impulsivity symptoms (Willcutt et al., 2000b; 2003). In contrast, Willcutt et al. (2003) found that bivariate linkage to chromosome 6p was somewhat stronger for reading difficulties and hyperactivity-impulsivity symptoms than reading difficulties and inattention symptoms. This sample was relatively small (N = 48 83 sibling pairs), however, and a much larger sample would be needed for this difference to reach statistical significance. Nonetheless, these inconsistent results suggest that the relations between reading difficulties and the DSM-IV ADHD symptom dimensions and subtypes are likely to be complex, and therefore warrant additional research. The neuropsychology of RD and ADHD Although several different types of studies are required to test definitively between all possible competing explanations for comorbidity (e.g., Neale & Kendler, 1995; Rhee et al., in press), Neuropsychological analyses of RD and ADHD 5 neuropsychological methods can be used to test several of the competing models described most frequently in the literature (e.g., Pennington et al., 1993; Rucklidge & Tannock, 2002; Willcutt et al., 2001). Therefore, in this section we briefly summarize current knowledge regarding the neuropsychology of RD and ADHD, then turn in the subsequent section to studies that used neuropsychological methods to examine the etiology of comorbidity between RD and ADHD. Neuropsychological profile of RD. Studies of individuals with and without reading difficulties suggest that phonological decoding, defined as the ability to translate sequences of printed letters into the corresponding sounds, plays a central role in both normal and abnormal reading development (e.g., Wagner & Torgesen, 1987; Pennington, 2002). The unique contribution of phonological decoding (PD) to most cases of RD has been suggested by the presence of significant group deficits in PD when older children with RD are compared to younger readers without reading disability who are reading at the same level (Olson, 1985; Rack, Snowling, & Olson, 1992). Moreover, twin studies have shown that there are strong genetic influences on the group deficit in PD (Gayán & Olson, 2001; Olson, Wise, Conners, Rack, & Fulker, 1989), and that these genes also influence the group deficit in word reading (Gayán & Olson, 2001; Olson, Forsberg, & Wise, 1994). Deficits in PD and word reading are in turn linked to genetic influences on deficits in the oral language skill of phoneme awareness, defined as the ability to recognize and manipulate the phonemic constituents of speech (Gayán & Olson, 2001; Olson et al., 1994). Problems with phoneme awareness are regarded by many as the most proximal cause of most cases of RD (c.f., Wagner, Torgesen, & Rashotte, 1994). Although deficits in groups with RD are most pronounced on measures of phoneme awareness and other facets of phonological processing, recent studies suggest that individuals with RD also have weaknesses in several other neurocognitive domains. These include difficulty accessing the orthographic representation of words from the lexicon (Gayán & Olson, 2001), weaknesses in other areas of speech and language processing (e.g., Pisecco et al., 2001; Olson, 1994), slower verbal naming speed (e.g., Compton, Olson, DeFries, & Pennington, 2002; Denckla & Rudel, 1976; Semrud-Clikeman, Guy, Griffin, & Hynd, 2000; Tannock, Martinussen, & Fritjers, 2000), and weaknesses in executive domains such as verbal working memory (e.g., Roodenrys, Koloski, & Grainger, 2001; Swanson, Mink, & Bocian, 1999; Willcutt et al., 2001), set-shifting (Weyandt, Rice, Linterman, Mitzlaff, & Emert, 1998), planning (e.g., Klorman et al., 1999), and response inhibition (e.g., Purvis & Tannock, 2000; Willcutt et al., 2001). Neuropsychological profile of ADHD. A large body of research indicates that groups with ADHD differ significantly from groups without ADHD on a variety of neurocognitive measures (see reviews by Barkley, 1997a, 1997b; Nigg, 2000, 2001; Pennington, 2002; Pennington & Ozonoff, 1996). Based on similarities between ADHD symptoms and the behavioral sequelae of frontal lobe injuries, several authors have proposed that ADHD is attributable to a core deficit in some facet of executive functions (EF), defined as cognitive functions that serve to maintain an appropriate problem-solving set in order to attain a future goal (e.g., Barkley, 1997a, 1997b; Pennington & Ozonoff, 1996). Previous theorists have criticized the construct of executive functions as weakly defined and overly broad (e.g., Pennington, 1997). Because many putative EF tasks are relatively nonspecific, poor performance on a specific task could be attributable to difficulties in any of several different aspects of the task. Moreover, definitions of EF often include a wide range of tasks, many of which appear to require somewhat different aspects of neurocognitive functioning. The multifactorial nature of EF is demonstrated by factor analyses of several batteries of putative EF measures (Mariani & Barkley, 1997; Miyake, Friedman, Emerson, Witzki, & Howerter, 2000; Pennington, 1997; Willcutt et al., 2001). All of these studies found that EF tasks tap more than one latent dimension of neurocognitive functioning. Although the specific factors varied somewhat among the studies, the overall pattern of results suggests that EF tasks may comprise at least four domains: (1) response inhibition, (2) working memory / updating, (3) setshifting / task-switching, and (4) interference control. Studies of children with a DSM-IV diagnosis of ADHD underscore the potential importance of the distinction between these different EF domains (Bedard et al., 2003; Chhabildas et al., 2001; Hinshaw, Carte, Sami, Treuting, & Zupan, 2002; Houghton et al., 1999; Klorman et al., 1999; Nigg, Blaskey, HuangPollock, & Rappley, 2002; Rucklidge & Tannock, 2002; Schmitz et al., 2002). A recent meta-analysis found that groups with DSM-IV ADHD performed significantly worse than groups without ADHD on measures of EF domains such as response inhibition, planning / organization, and working memory, as well as measures of domains with less of an executive component such as processing speed, rapid naming, and fine and gross motor skill (Willcutt et al., 2004). In contrast, the ADHD groups were not consistently impaired on measures of set-shifting or interference control. Neuropsychological analyses of RD and ADHD 6 Although these studies suggest that ADHD is associated with significant weaknesses in several EF domains, these results challenge the hypothesis that any specific EF deficit is the core neurocognitive deficit in ADHD. The mean effect size between groups with and without ADHD is moderate for each of the EF measures (d = 0.4 – 0.6; Willcutt et al., 2004), suggesting that none of these neurocognitive weaknesses is a necessary or sufficient cause of ADHD. Moreover, many of these studies did not assess important covariates such as IQ and reading achievement, leaving open the possibility that the EF deficits associated with ADHD could be attributable to group differences in IQ or to the association between ADHD and RD (Pennington & Ozonoff, 1996). Thus, in contrast to the consistently strong relation between phonological processing weaknesses and RD, the neuropsychological profile of ADHD is not as well understood, and the core neuropsychological deficit remains elusive. Using neuropsychology to understand comorbidity To fully understand the neuropsychological correlates of RD and ADHD and the etiology of comorbidity between the two disorders, groups with RD only, ADHD only, and both RD and ADHD must be directly compared on the same measures. The common etiology, phenocopy, cross-assortment, and cognitive subtype hypotheses each make several key predictions regarding the relations among the groups on these measures. Predictions of the competing hypotheses regarding a double dissociation between RD and ADHD. A double dissociation occurs when two disorders are associated with opposite patterns of impairment in two different cognitive domains. Because the phenocopy, cross-assortment, and cognitive subtype hypotheses suggest that RD and ADHD are caused by different etiological influences, all three models predict a significant double dissociation between RD and ADHD. In contrast, although the common genetic etiology hypothesis also predicts a double dissociation between RD and ADHD for all measures that are not influenced by the common genetic effects, it predicts that both the RD-only and ADHD-only groups will exhibit weaknesses in any neuropsychological domains that are influenced by the common genes. In the first study that used a full 2 X 2 (RD X ADHD) design to examine performance on measures of EF and phonological processing, Pennington et al. (1993) found a double dissociation between RD and ADHD. Specifically, the group with ADHD alone was significantly impaired on EF measures but not measures of phonological processing, whereas the group with RD alone exhibited severe phonological processing deficits but was not impaired on the EF measures. Subsequent studies generally supported this double dissociation, although several suggested that RD may also be associated with mild deficits on at least a subset of EF measures (e.g., Klorman et al., 1999; Rucklidge & Tannock, 2002; Nigg et al., 1998; Purvis & Tannock, 2000; Shaywitz et al., 1995; Willcutt et al., 2001). These mixed results provide some support for all four hypotheses, suggesting that additional research is needed to better understand which neurocognitive weaknesses are specific to RD or ADHD and which neurocognitive difficulties are associated with both disorders. Predictions of the competing hypotheses regarding the comorbid group. Each competing hypothesis makes a different set of predictions regarding the neuropsychological profile of the comorbid group in relation to the groups with RD and ADHD alone. The phenocopy hypothesis suggests that one disorder often causes the symptoms of the second disorder in the absence of the neurocognitive weaknesses associated with the second disorder when it occurs alone. Therefore, it predicts that the comorbid group will exhibit the same pattern of neuropsychological deficits that are present in the group with the first disorder alone. The cognitive subtype hypothesis is less specific; it simply predicts a significant RD X ADHD interaction on at least some measures, such that the neuropsychological deficits of the comorbid group differ from the simple additive combination of the deficits associated with RD and ADHD when they occur alone. Because the common genetic etiology and cross-assortment hypotheses each suggest that individuals with comorbid RD+ADHD have the risk factors for both RD and ADHD, these models each predict that the comorbid group will exhibit the neuropsychological weaknesses that are associated with each disorder when it occurs alone. In contrast, the common etiology hypothesis predicts that all three clinical groups will exhibit weaknesses on at least some of the same neuropsychological tasks, whereas the cross-assortment hypothesis predicts that the comorbid group will exhibit the additive combination of the distinct neuropsychological weaknesses that are associated with RD and ADHD. Initial results suggested that the comorbid group exhibited significant phonological processing deficits in the absence of the EF deficits associated with ADHD when it occurred in the absence of RD Neuropsychological analyses of RD and ADHD 7 (Pennington et al., 1993), but most later studies found that the comorbid group exhibited both the EF deficits associated with ADHD and the phonological weaknesses associated with RD (e.g., Nigg et al., 1998; Willcutt et al., 2001). Some recent studies have even suggested that comorbid RD may be a marker for a group of children with ADHD with more severe cognitive deficits (e.g., Purvis & Tannock, 2000; Seidman et al., 2001; Willcutt et al., 2001). Taken together, these results provide minimal support for the phenocopy hypothesis, but suggest that additional research is needed to test the common genetic etiology, cross-assortment, and cognitive subtype hypotheses. The present study An extensive battery of neuropsychological measures was administered to groups with ADHD-only, RD-only, ADHD+RD, and neither ADHD nor RD. The sample described in this paper is completely independent from the sample used in our previous paper (Willcutt et al., 2001). The primary goals of the study were as follows: 1) An exploratory factor analysis of the neurocognitive battery was conducted to examine the relations among the neuropsychological variables and simplify interpretation. We predicted that the battery of EF measures would be best described by two or more latent factors, rather than a single factor indicating that the EF tasks tap a unitary construct. 2) The neuropsychological profile of groups with RD, ADHD, RD+ADHD, and neither RD nor ADHD were compared to clarify the pattern of neurocognitive weaknesses associated with RD and ADHD independent of the influence of the other disorder. Based on previous research, it was predicted that RD would be associated with prominent weaknesses on all measures of component reading and language skills, coupled with milder weaknesses on EF and processing speed tasks. In contrast, we anticipated that ADHD would be associated with deficits on EF and processing speed tasks that would be most pronounced on measures of response inhibition, but that ADHD would not be independently associated with deficits on the reading and language measures. 3) To examine the etiology of comorbidity between RD and ADHD, the neuropsychological profile of the three clinical groups was compared. Based on our previous results and other findings in the literature, we predicted that the comorbid group would exhibit the deficits associated with both RD and ADHD, providing further evidence against the phenocopy and cognitive subtype hypotheses as explanations for comorbidity between RD and ADHD. 4) By examining the extent to which each neuropsychological deficit was associated with the three clinical groups, we tested the potential utility of these neuropsychological measures as markers for the common genetic etiology of RD and ADHD. Although we did not have strong a priori predictions for this analysis, some previous studies reported promising results for measures of processing speed (e.g., Rucklidge & Tannock, 2002). 5) Finally, a series of ancillary analyses were conducted to test if the pattern of neuropsychological weaknesses varied as a function of ADHD subtype. Based on initial results in this sample (Chhabildas et al., 2001), we did not anticipate any significant differences between the inattentive and combined subtypes.