How to Cite this Article: Sheikhi AR, Martin N, Hay D, Piek JP. 2012. Phenotype Refinement for Comorbid Attention Deficit Hyperactivity Disorder and Reading Disability. Am J Med Genet Part B 162B:44–54.
Phenotype refinement for comorbid attention deficit hyperactivity disorder and reading disability†
Article first published online: 29 NOV 2012
Copyright © 2012 Wiley Periodicals, Inc.
American Journal of Medical Genetics Part B: Neuropsychiatric Genetics
Volume 162, Issue 1, pages 44–54, January 2013
How to Cite
Sheikhi, A. R., Martin, N., Hay, D. and Piek, J. P. (2013), Phenotype refinement for comorbid attention deficit hyperactivity disorder and reading disability. Am. J. Med. Genet., 162: 44–54. doi: 10.1002/ajmg.b.32119
- Issue published online: 18 DEC 2012
- Article first published online: 29 NOV 2012
- Manuscript Accepted: 5 NOV 2012
- Manuscript Received: 11 APR 2012
- National Health and Medical Research Council (NHMRC). Grant Number: 229005
- latent class analysis;
- Top of page
- MATERIALS AND METHODS
Comorbidity between Attention Deficit Hyperactivity Disorder (ADHD) and reading disability (RD) is common; however, the heritability of this comorbidity is not well understood. This may be due to the complexity and heterogeneity of ADHD and RD phenotypes. Using alternative ADHD–RD sub-phenotypes instead of those arising from the DSM-IV may lead to greater success in the search for comorbid ADHD–RD susceptibility genes. Therefore, this study aims to refine ADHD–RD phenotypes into homogenous informative sub-phenotypes using latent class analysis (LCA). LCA was performed on 2,610 Australian twin families (6,535 individuals) in order to generate probabilistic genetically distinct classes that define ADHD–RD subtypes, including comorbidity, based on related symptom clusters. The LCA separated the phenotypes for ADHD and RD into nine classes. One class was unaffected; three classes demonstrated the three DSM-IV subtypes of ADHD, three subtypes showed different severities of RD, and two classes expressed a combination of RD and ADHD subtypes. LCA proved effective in refining the phenotypes of ADHD alone, RD alone, and ADHD–RD comorbidity, and its ability to classify them into homogenous groups based on clusters of symptoms, suggesting that the latent classes may be robust enough to use in molecular genetic studies. © 2012 Wiley Periodicals, Inc.
- Top of page
- MATERIALS AND METHODS
Attention Deficit Hyperactivity Disorder (ADHD) is a complex neuro-developmental disorder as classified in the Diagnostic and Statistical Manual of Mental Disorders, fourth edition DSM-IV-TR [APA, 2000]. ADHD symptom-based phenotypes are distinguished by the presence of developmentally inappropriate levels of impulsivity, hyperactivity, and inattentiveness [Stefanatos and Baron, 2007]. The classification and phenotypes of ADHD are still uncertain [Levy and Hay, 2001]. Several quantitative genetic studies have revealed that genetic factors are the major influence within family susceptibility, with heritability estimates for ADHD ranging from 60% to more than 90% [Levy et al., 1997; Biederman and Faraone, 2005; Wood et al., 2010]. This indicates that specific genes play a role in the aetiology of ADHD.
ADHD symptoms tend to change with development. For example, Barkley  found that Hyperactive–Impulsive symptoms appeared earlier (3–4 years old) than Inattentive symptoms, which started around school age (5–7 years old). Also, Hyperactive–Impulsive symptoms gradually decrease with age and development. Hay et al.  believed that the presence of common environmental factors such as family, school or medication intervention may eventually influence the symptoms of the Hyperactive–Impulsive ADHD. Cohen et al.  stated that ADHD is less prevalent in the younger age groups, with the exception of the Inattentive subtype in females, which can increase as they get older [Levy et al., 2005].
Reading disability (RD) is also a complex neurobehavioral disorder that affects approximately 5–10% of school-aged children, irrespective of intelligence, education, and social environment [Shaywitz et al., 1990; Meng et al., 2011]. A person impaired with RD cannot interpret written words and would find it difficult or impossible to spell and decode words as a consequence of a deficiency in language phonology. Phenotypically, RD has been defined by partitioning reading skill development into its major contributing cognitive components; namely, phonological awareness, phonological coding, orthographic coding, and rapid serial naming [Zumberge et al., 2007; Pennington, 2009]. Behavioral and molecular genetic studies on each of these cognitive components individually have indicated that they can be successfully used in isolation as RD phenotypes. These results support the idea that the specific cognitive components involved in reading skill development may each map neatly to specific genomic regions [McGrath et al., 2006].
Comorbidity between ADHD and RD is common, and they co-occur significantly more frequently than would be expected by chance because of a phenotypic overlap [Pennington, 2006; Cheung, 2012]. Despite the current advances in ADHD and RD genetic studies, the susceptibility genes for ADHD–RD comorbidity remain uncertain; this might be due to the complexity and heterogeneity of ADHD and RD phenotypes. Therefore, using alternative ADHD–RD sub-phenotypes, instead of those arising from the DSM-IV [APA, 2000], may lead to more rapid success in the search for ADHD–RD comorbidity susceptibility genes.
Despite the phenotypic characterizations of ADHD, defined through DSM-III [APA, 1980], DSM-III-R [APA, 1987], DSM-IV [APA, 1994], and DSM-IV-TR [APA, 2000], the etiology of ADHD is still unclear. The validity of the disorder, because of its co-occurrence with other disorders, and the suitability of dimensional versus categorical models are debatable. Although DSM editions III through IV-TR have defined ADHD using categorical phenotypes which have marked heritability, the use of these categories has potential problems for genetic analyses. For example, an individual can have 10 symptoms of DSM-IV ADHD, but if five are Inattentive and five are Hyperactive–Impulsive the individual is classified as unaffected. Another person may have 10 symptoms of ADHD, six Inattentive and four Hyperactive–Impulsive, and would be categorized as having the Inattentive subtype, while yet another individual with ten symptoms of ADHD, four Inattentive and six Hyperactive–Impulsive, would be categorized as Hyperactive–Impulsive.
Phenotypes based on DSM-IV criteria have not been very successful as reference points in molecular genetic studies as they are heterogeneous and therefore unable to detect the susceptible gene(s) contributing to these particular phenotypes [Georgiades et al., 2007]. This may be one of the reasons for the delay in the identification of genes linked to ADHD–RD comorbidity, as the most genetically informative phenotypes have not been fully identified. Khan and Faraone  suggested that DSM-IV diagnostic criteria for ADHD do not focus on its complexity, heterogeneity, and comorbidity, but only considers grouping symptoms. As a result, there is no distinct line between symptoms of ADHD and symptoms involving its comorbidity with RD. This makes it difficult to decide if symptoms of ADHD represent more than one disorder or whether these symptoms represent distinct subtypes of ADHD [Volk et al., 2006].
Another obstacle is that defining a child with DSM-IV comorbid ADHD–RD requires two sets of arbitrary cut-offs, which may lead to an inappropriate classification, making the sample heterogeneous. Todd and his research team demonstrated in several studies [e.g., Hudziak et al., 1998; Neuman et al., 1999; Rasmussen et al., 2002b; Todd et al., 2002, 2005; Volk et al., 2006] the efficiency of using latent class analysis (LCA) to obtain homogenously distinct groups that are appropriately classified, as LCA has the advantage of identifying naturally occurring clusters of symptoms without the need for symptom number cut-offs [Volk et al., 2006]. Hudziak et al.  successfully applied the pioneer work of McCutcheon  on LCA using DSM-IV ADHD symptoms to determine whether different types of impairment (academic, peer or family) are differentially associated with those subtypes, and also to determine whether the distribution of ADHD symptomatology is more consistent with dimensional or categorical models. The results showed at least eight classes were needed to account for the 926 unique symptom profiles reported by parents. Therefore, Neuman et al.  encouraged the adoption of LCA in molecular genetic studies of ADHD as symptom clusters identified through LCA seem to be a more suitable approach for such studies than symptoms based on DSM-IV diagnostic criteria.
LCA can be used to cluster each group into subphenotypes based on homogenous symptomology. These subphenotypes can then be utilized in a logistic regression-based analysis to determine odds ratios independently, for MZ versus DZ concordance, both within-subtype and between-subtype heritability. The results suggest that the latent class approach may be useful in studying the genetics of ADHD, particularly in enabling a molecular genetic approach to determining the loci relevant to the aetiology and expression of symptoms. LCA has the ability to place individuals into phenotypically similar groups to produce distinctive and heritable classes [Volk et al., 2006].
Heterogeneity among DSM-IV ADHD subtypes is an obstacle to overcome when examining the distinction between the subtypes. Szatmari et al.  argued that when identifying psychiatric disorders such as ADHD, the DSM-IV is problematic as the number and severity of symptoms is so variable. Accordingly, an alternative approach to identifying candidate ADHD genes may be to refine the ADHD phenotypes so that there would be an “informative phenotype” for performing genetic analysis. Szatmari et al.  argued that an informative phenotype would be more Mendelian-like and could be transmitted within the pedigree in less complex ways, not like the DSM-IV phenotype. The informative phenotype can be categorized into component, intermediate and covariate phenotypes [Szatmari et al., 2007]. Because ADHD is a complex disorder influenced by multiple genes, and each ADHD subtype has a wide range of heterogeneous phenotypes controlled by different genetic mechanisms, this study proposed that component phenotypes can effectively describe the fundamental characteristics of each subtype, instead of selecting a DSM-IV subtype that contains a wide range of phenotypes. LCA can identify component phenotypes, as this analysis has the capability of segregating ADHD phenotypes into appropriate symptom clusters and to re-classify the heterogeneous ADHD phenotypes into distinctive homogenous groups, producing valid genetically informative phenotypes. Hence, this study aimed to utilize LCA to produce more homogenous, genetically informative phenotypes to identify ADHD alone, RD alone, and ADHD–RD.
MATERIALS AND METHODS
- Top of page
- MATERIALS AND METHODS
The participants for this study came from twin families who participated in the fourth wave of the Australian Twin ADHD Project (ATAP) [Bennett et al., 2006]. These twin families were members of the Australian Twin Registry, which is a nationwide volunteer-based twin registry having approximately 30,000 pairs of twins. At the time of this study, ATAP had recruited 2,610 twin families out of 3,500 twin families approached (a response rate of 75%). These families constituted the participants of this study, including monozygotic (MZ), dizygotic (DZ) twins, and their siblings. The 2,610 twin families gave a total twins and siblings number of 7,209. The total number of male children was 3,681 (51.1%), and that of female children was 3,528 (48.9%). There were 2,262 (31.4%) monozygotic twins and 2,910 (40.4%) dizygotic twins, with 25 families excluded as they had no zygosity information. The total number of siblings was 2,039, 1,609 (22.3%) as sibling 1 and 430 (6.0%) as sibling 2. The age range of the total sample was from 4 to 18 years old, with a mean age of 12.94, ±3.9 years.
This study applied the exclusionary criteria that Hay et al.  established on the four waves of ATAP. The criteria for exclusion involved any one of the following problems: if any of the twins or siblings suffered from any mental retardation, psychosis, autism or a major medical or neurological illness including deafness, blindness, cerebral palsy, as well as major cardiac malformations. Children with obvious physical or health problems, such as retinopathy, were thus excluded [Hay et al., 2002]. Other conditions such as muscular dystrophy, leukemia, Down's syndrome or rare genetic conditions or specific environmental disorders such as meningitis were also causes for exclusion. The exclusion extended to the whole family; if one child suffered from a disability or other identified developmental disorders other than ADHD or RD all family members were excluded from the study. The criteria also excluded families who participated previously in long-term behavioral studies. Another criterion for exclusion was multiples who were born both pre-term and with extremely low birth weight, given the likelihood of subsequent difficulties as a result of this.
The Australian Twin Behaviour Rating Scale (ATRBS)
The ATRBS was designed by Levy et al. [1996, 1997] and used with the first wave of the ATAP. It was originally based on DSM-III-R, and subsequently DSM-IV, and designed to measure the presence of childhood behavioral disorders such as ADHD, Reading and Spelling Disorder. Levy et al. [1996, 1997] stated that the use of the ATBRS based on parent ratings can be considered as a conservative sign of symptoms present. Levy et al.  found the criteria of DSM-III-R ADHD symptoms, speech and language problem symptoms, and reading disability symptoms, were highly reliable (0.86, 0.71, and 0.82, respectively) based on Cronbach's alpha. The ATBRS was refined and renamed the “Twin and Sibling Questionnaire.”
Because this study sought to investigate DSM-IV ADHD subtypes and ADHD comorbidity with reading disability (RD), only certain information was used: the DSM-IV ADHD and reading disability symptoms of the twins and their siblings, plus demographic data such as sex, age, and zygosity.
Hay et al.  designed 14 questions to assess zygosity. Six questions look at similarity of features and six questions focus on confusion of the twins' identities [Nichols and Bilbro, 1966; Cohen et al., 1975]. The other two questions were used to identify placentation and blood group polymorphisms in order to distinguish MZ twins. Determining the zygosity based on placentation alone is not 100% accurate, as about one-third of MZ twin pairs can have two separate placentas.
DSM-IV ADHD items measure
The Twin and Sibling Questionnaire [Hay and Levy, 2004] contains 18 items based on the ADHD symptoms listed in the DSM-IV, nine items for Inattention, six items for Hyperactivity, and three items for Impulsivity. The Twin and Sibling Questionnaire uses a four-point scale code. For each item, responses could range from, “not at all” (scored as 0) to, “very much/often” (scored as 3). This method of establishing symptom endorsement is a valid way of identifying children with subtypes of ADHD [Levy and Hay, 2001]. A parent rating of 0 or 1 means that symptoms were absent, and a parent rating of 2 or 3 means symptoms were present.
Identification of participants was based on DSM-IV ADHD diagnostic criteria. In order for a child to be diagnosed with Inattentive or Hyperactive–Impulsive ADHD, he or she should have six or more symptoms out of nine for either subtype respectively. For a child to be diagnosed with the Combined subtype, he or she should have six or more Inattentive symptoms, and six or more Hyperactive–Impulsive symptoms. In addition, the Inattentive and Hyperactive–Impulsive scores were calculated based on the sum of nine scaled items for Inattention and the sum of nine scaled items for Hyperactivity–Impulsivity, which gave a maximum score of 27 for each domain, whereas the scores of the Combined domain were calculated based on the sum of nine scaled items for both Inattention and Hyperactivity–Impulsivity, giving a maximum score of 54. Cronbach's Alpha for the Inattentive items was 0.857 and for the Hyperactive–Impulsive items, 0.825.
Reading disability measure
The Twin and Sibling Questionnaire contains seven items measuring RD. This was originally created and designed by Erik Willcutt and his colleagues in the “Learning and Behavior Questionnaire” for the Colorado Learning Disabilities Research Center (CLDRC) [Willcutt et al., 2003a]. In 2003 they tested the internal and external validity and reliability of the seven items, which showed high validity and reliability. In addition, the correlations were 0.65, 0.61, 0.61, and 0.71 when tested with the Peabody Individual Achievement Tests (PIAT) for Reading Recognition, Reading Comprehension, and Spelling and Reading Composite scores respectively.
Willcutt et al.  reassessed the validity and reliability of the seven reading items and found that only six loaded strongly on the first rotated factor when exploratory and confirmatory factor analyses were performed on four different groups. One item, “Difficulty learning days and months,” exhibited a weak loading in all four samples. The internal consistency represented by Cronbach's alpha ranged from 0.89 to 0.94 for the six items' composite score. The analyses also showed high correlations for inter-rater (r = 0.84) and test-retest (r = 0.82) reliability. The Learning and Behavior Questionnaire's (LBQ) results also exhibited significant correlations for single word reading measure (r = 0.61–0.71), reading fluency (r = 0.41–0.55), and reading comprehension (r = 0.42–0.58). As expected, the correlations testing discriminate validity were low ranging from 0.15 to 0.40 with intelligence, short- and long-term verbal memory, motor functioning, and math achievement.
The assessment of participants with RD in the current study was based on the earlier version with seven RD items [Willcutt et al., 2003a], as the assessments were completed prior to the publication of the more recent version. In the current study, the seven RD items had high internal consistency with Cronbach's alpha ranging from 0.913 to 0.928 and high inter-item correlations ranging from 0.688 to 0.840. The scores on each RD item were added together to produce a total RD score for each twin and sibling. The seven RD items offered a continuous measure, giving a maximum score of 21. Any child who gained a score of seven or more was defined as “RD affected,” whereas any child with a score of <7 was defined as “RD unaffected.”
Ethical approval for this study was obtained from both the Curtin University Human Research Ethics Committee and the ethics committee of the Australian Twin Registry. Each family was mailed a package containing an information sheet about the study, and a parent-report questionnaire entitled “Twin and Sibling Questionnaire” [Hay and Levy, 2004]. The completed questionnaire packages were mailed back to Curtin University in “Reply Paid” envelopes. Each returned questionnaire contained a detachable personal information sheet for each twin family which were kept in a secure cabinet, separate from the completed ATR-labelled questionnaire in order to ensure confidentiality.
The data for the LCA were based on parent responses about their offspring using the 18 DSM-IV defined ADHD items and seven RD items [Willcutt et al., 2003a]. The LCA Program (LCAP) Version 2.34 [Neuman et al., 1999] was used to investigate the observed association among the variables. Accordingly, the form of data used in LCA is often categorical, resulting in the identification of distinct diagnostic subtypes that can then be used as a classification tool. The maximum likelihood algorithm (EM algorithms) was utilized [Dempster et al., 1977] in the LCAP.
As the parents' endorsement of the DSM-IV ADHD and RD items [Willcutt et al., 2003a] for their children was based on a four-point scale, these data were implemented in LCAP in a categorical form; “zero” and “one” coded as “zero” meaning “unaffected,” and “two” and “three” were coded as “one” meaning “affected.” These categorical data were fitted to latent class solutions from one to sixteen in LCAP [Neuman et al., 1999] by a maximum likelihood algorithm (EM algorithm) [Dempster et al., 1977]. In addition, three elements were considered for estimating the best fitting model: a Bayesian Information Criterion (BIC), the likelihood ratio Chi square (χ2), and class membership criteria stability. As the number of those 16 classes increased (1–16), the Chi square goodness-of-fit gradually improved which can be indicated by a decline in the BIC. The best-fitting solution was based on the lowest BIC and was considered as the best representative of the sample.
The subtypes of the best latent class solutions were marked by a line chart based on the pattern of symptom endorsements, which represented probabilities for each class in the LCA output file. The line chart was used to show the strength and/or weakness for each ADHD and RD symptom amongst the sample.
The results section describes characteristics of the ADHD/RD latent classes, DSM-IV ADHD subtypes, and RD categories. These characteristics include prevalence, sex, and age differences for each criterion. In addition, we also describe the overlap between the ADHD–RD latent classes and DSM-IV ADHD subtypes in the total sample, and in the male and female samples. The extent to which both ADHD–RD latent class criteria and DSM-IV ADHD subtypes are compatible is determined by examining the degree of overlap to see if individuals are assigned to the same or different phenotype using both classification criteria. The characteristics of ADHD–RD comorbidity using both sets of criteria is also described, giving the prevalence and sex differences between the two sets of criteria. Furthermore, this analysis highlights the prevalence of MZ and DZ twins assigned to each ADHD–RD latent class, compares and estimates the expression of DSM-IV ADHD and RD item endorsements in DSM-IV ADHD subtypes, the RD category, and ADHD–RD latent classes.
- Top of page
- MATERIALS AND METHODS
The LCA results showed that there were two best-fitting models, one with eight latent classes (LC-8) and one with nine (LC-9). This finding was based on selecting the lowest Bayes Information Criteria (BIC) and improved likelihood ratio Chi square. In this instance, it was found that the BICs for both LC-8 (77887.94269) and LC-10 (77881.10337) were higher than the BIC for LC-9 (77858.02140). Accordingly, as LC-9 was the lowest, it was selected as the best fitting model to represent the data (see Fig. 1).
Characteristics of the Nine ADHD/RD Latent Classes
The prevalence of the nine ADHD/RD latent classes
The latent class-9 model exhibited nine ADHD–RD latent classes. The prevalence of each of these is listed in Table I. One class was “Few Symptoms,” three classes each similar to the three subtypes of ADHD, three subtypes showing different severities of RD, and two classes expressing comorbidity of RD with two different ADHD subtypes.
|Total Latent Class-9||Frequency||Total sex|
|1||Few Symptoms||4,422 (67.7%)||2,038 (31.2%)||2,384 (36.5%)|
|2||Predominantly Hyp–Imp||502 (7.7%)||281 (4.3%)||221(3.4%)|
|3||Moderate RD||362 (5.5%)||215 (3.3%)||147 (2.2%)|
|4||Predominantly inattentive||389 (6.0%)||243 (3.7%)||146 (2.2%)|
|5||Severe RD||233 (3.6%)||134 (2.1%)||99 (1.5%)|
|6||Predominantly inattentive & RD||182 (2.8%)||126 (1.9%)||56 (0.9%)|
|7||Combined||180 (2.8%)||128 (2.0%)||52 (0.8%)|
|8||Unique Severe RD||147 (2.2%)||97 (1.5%)||50 (0.8%)|
|9||Combined & RD||118 (1.8%)||87 (1.3%)||31 (0.5%)|
|Total||6,535 (100%)||3,349 (51.2%)||3,186 (48.8%)|
Sex differences among the nine ADHD–RD latent classes
There were more affected males than females in all categories (20.1% total affected males vs. 12.3% total affected females) (see Table I). There were also sex differences in the prevalence of the different ADHD subtypes. For males, inattention was more prevalent (total for “Predominantly Inattentive” and “Predominantly Inattentive & RD” was 5.6%) than “Predominantly Hyperactive–Impulsive” (4.3%). The opposite was true for females (total for “Predominantly Inattentive” and “Predominantly Inattentive & RD” is 3.1% vs. 3.4% for “Predominantly Hyperactive–Impulsive”) (see Table I).
The Overlap Between ADHD/RD Latent Class-9 and DSM-IV ADHD Subtypes
Table II shows the overlap between the nine ADHD–RD latent classes and the DSM-IV ADHD subtypes across the sample. Although classes refined by LCA exhibited major differences with DSM IV categories, Table II shows overlapping of DSM-IV ADHD subtypes with the ADHD/RD latent classes (95.1% “No ADHD,” 2.7% Inattention, 1.0% Hyperactive–Impulsive). The DSM-IV identified 2.7% as Inattentive cases; 1.0% as Hyperactive–Impulsive, and 1.2% as Combined subtypes. However, the LCA identified 6.0% as “Predominantly Inattentive,” 7.7% as “Predominantly Hyperactive–Impulsive,” and 2.8% as “Combined” latent classes. Only 7.2% of Inattentive, 2.6% Hyperactive–Impulsive, and 14.4% of combined cases showed overlap between the two categories.
|9-ADHD/RD Latent Classes||DSM-IV ADHD subtypes||Total|
|Few Symptoms||4,375 (98.9%)||31 (0.7%)||9 (0.2%)||7 (0.2%)||4,422 (67.7%)|
|Predominantly Hyp–Imp||478 (95.2%)||7 (1.4%)||13 (2.6%)||4 (0.8%)||502 (7.7%)|
|Moderate RD||354 (97.8%)||4(1.1%)||3 (0.8%)||1 (0.3%)||362 (5.5%)|
|Predominantly Inattentive||351 (90.2%)||28 (7.2%)||8 (2.1%)||2 (0.5%)||389 (6.0%)|
|Severe RD||227 (97.4%)||3 (1.3%)||2 (0.9%)||1 (0.4%)||233 (3.6%)|
|Predominantly Inattentive & RD||147 (80.8%)||25 (13.7%)||4 (2.2%)||6 (3.3%)||182 (2.8%)|
|Combined||100 (55.6%)||34 (18.9%)||20 (11.1%)||26 (14.4%)||180 (2.8%)|
|Unique Severe RD||141 (95.9%)||5 (3.4%)||0 (0.0%)||1 (0.7%)||147 (2.2%)|
|Combined & RD||40 (33.9%)||41 (34.7%)||8 (6.8%)||29 (24.6%)||118 (1.8%)|
|Total||6,213 (95.1%)||178 (2.7%)||67 (1.0%)||77 (1.2%)||6,535 (100%)|
Endorsement of RD items among the RD latent classes
Out of the nine latent classes (Table I), three included individuals with RD but no ADHD symptoms. These were described as “Moderate RD” (with a prevalence of 5.5%), “Severe RD” (3.6%), and “Unique Severe RD” (2.2%). As can be seen from Table III, the average endorsement probabilities of the seven RD items [Willcutt et al., 2003a] with the Moderate RD (0.33) were the lowest, compared to the “Severe RD” (0.65) and “Unique Severe RD” (0.93).
|RD Latent Class item endorsement probabilities|
|Moderate RD||Severe RD||Unique Severe RD|
RD items [Willcutt et al., 2003a
|(n = 362) 5.5%||(n = 233) 3.6%||(n = 147) 2.2%|
|1. Difficulty with spelling||0.70||0.81||0.97|
|2. Difficulty learning letter names||0.25||0.17||0.90|
|3. Difficulty learning phonics||0.45||0.70||1.00|
|4. Slow reading more than other children of the same age||0.27||1.00||0.92|
|5. Reading below expectancy level||0.07||1.00||0.88|
|6. Difficulty learning the days or months||0.07||0.10||0.87|
|7. Extra help in school with problems in reading or spelling.||0.52||0.76||1.00|
Examining ADHD–RD latent classes with highly endorsed phenotypic RD items
Table IV shows the correspondence between the RD latent classes and the RD components: spelling, verbal learning and memory, phonological awareness, rapid memory, and overall reading ability. Upon examination of the RD items in each RD and ADHD–RD latent class, it was found that RD item 1 (spelling), was highly endorsed in the “Moderate RD” latent class. The most highly endorsed of the seven RD items with the “Severe RD” latent class were rapid memory, the overall reading ability, and spelling. However, all seven RD items were found to be highly endorsed with the “Unique Severe RD” latent class. With the “Inattentive-RD” latent class, the predominant RD items were spelling, rapid memory item, and overall reading ability, whereas, with the “Combined RD” latent class, all RD items were highly endorsed.
|RD Latent Class||Highly endorsed phenotypic RD item(s)|
|1. Moderate RD latent class||Item 1(spelling)|
|2. Severe RD||Item 4 (rapid memory)|
|Items 5 and 7 (overall reading ability)|
|Item 1 (spelling)|
|3. Predominantly Inattentive RD||Item 1 (spelling)|
|Item 4 (rapid memory)|
|Items 5 and 7 (overall reading ability)|
|4. Unique Severe RD||Item 1 (spelling)|
|Items 2 and 6 (verbal learning and memory)|
|Item 3 (phonological awareness)|
|Item 4 (rapid memory)|
|Items 5 and 7 (overall reading ability)|
|5. Combined RD||Item 1 (spelling)|
|Items 2 and 6 (verbal learning and memory)|
|Item 3 (phonological awareness)|
|Item 4 (rapid memory)|
|Items 5 and 7 (overall reading ability)|
Characteristics of ADHD/Reading Disability Comorbidity
Prevalence of the nine ADHD–RD latent class comorbidity
The total comorbidity of the nine ADHD–RD latent classes with the “Yes RD” was 14.9%. Furthermore, the comorbidity of the “Moderate RD,” the “Severe RD,” and the “Predominantly Inattentive/RD” latent classes with “No RD” category was 1.9%, 0.6%, and 0.2%, respectively. These rates revealed the accuracy of LCA in grouping the related phenotypes together.
The LCA showed both the “Predominantly Hyperactive–Impulsive” (0.2%) and “Combined” (0.2%) latent classes were the lowest classes that were comorbid with RD. On the other hand, the “Combined-RD” latent class (1.8%) was moderately high, but not as high as the “Predominantly Inattentive-RD” (3.6%) class.
Characterization of the Nine ADHD–RD Latent Classes With Zygosity
The traditional method to determine heritability is by comparing concordance rates of MZ twins versus DZ twins [Tishler and Carey, 2007]. Heath et al.  stated that LCA can be considered an alternative statistical method for determining zygosity based on the zygosity questionnaire items. LCA is also considered an effective tool for determining zygosity when genotyping information is not present [Heath et al., 2003]. Todd et al.  stated that by comparing the degree of MZ and DZ concordance between twins in the same and different latent classes, researchers can determine if these latent classes are heritable or not and also can recognize a genetic influence on that particular latent class. Instead of using an odds ratio test, our study used the Chi-square test.
Table V shows cross sub-phenotype heritability by calculating the Chi-square test for the proportion of concordant versus discordant pairs between MZ and DZ twins. According to this, when the Chi-square test was calculated for the concordant twins, it showed a significant difference between MZ and DZ (χ2 = 50.104; P < 0.01), which indicates that those latent classes are distinctive heritable groups. As can be seen in Table V, one exception was the number of concordant MZ twins for the “Predominantly Inattentive-RD” latent class compared to DZ twins, where there were more DZ children than MZ children (11 and 10 respectively). Similarly, Table V shows how the proportion of concordant versus discordant pairs differ between MZ and DZ twins. The Chi-square test showed significant differences between concordant and discordant MZ and DZ twins for all nine classes except the “Predominantly Inattentive-RD” class. These significant results indicated that these latent classes are heritable.
|T1 × T2 Latent Class||Zygosity||Concordant twins||Discordant twins||Chi square|
|1. T1 × T2 Few Symptoms||MZ||665 (64.5%)||48 (4.8%)||χ2 = 76.83, P < 0.01|
|DZ||679 (51.5%)||200 (15.2%)|
|2. T1 × T2 Predominantly Hyp–Imp||MZ||49 (4.8%)||33 (7.7%)||χ2 = 24.45, P < 0.01|
|DZ||24 (1.8%)||77 (5.9%)|
|3. T1 × T2 Moderate RD||MZ||42 (4.1%)||18 (1.8%)||χ2 = 41.46, P < 0.01|
|DZ||11 (0.8%)||62 (4.8%)|
|4. T1 × T2 Predominantly Inattentive||MZ||29 (2.8%)||25(2.5%)||χ2 = 24.13, P < 0.01|
|5. T1 × T2 Severe RD||MZ||23(2.2%)||14 (1.4%)||χ2 = 22.39, P < 0.01|
|DZ||7 (0.5%)||44 (3.4%)|
|6. T1 × T2 Predominantly inattentive & RD||MZ||10 (1.0%)||17 (1.7%)||χ2 = 0.913, P > 0.05|
|DZ||11 (0.8%)||31 (3.2%)|
|7. T1 × T2 Combined||MZ||13 (1.3%)||5 (0.5%)||χ2 = 23.88, P < 0.01|
|DZ||2 (0.2%)||30 (2.3%)|
|8. T1 × T2 Unique severe RD||MZ||13 (1.3%)||5 (0.5%)||χ2 = 14.15, P < 0.01|
|DZ||8 (0.6%)||31 (2.3%)|
|9. T1 × T2 Combined & RD||MZ||14 (1.4%)||8 (0.8%)||χ2 = 6.42, P < 0.05|
|DZ||6 (0.5%)||17 (1.3%)|
The Overlapping and Distinct Cases Between the Nine ADHD–RD Latent Classes and the DSM-IV ADHD Subtypes by “No RD” and “Yes RD” Categories
Table VI shows the overlap between the nine ADHD–RD latent classes and the DSM-IV ADHD subtypes by “No RD” and “Yes RD” categories. This confirms the ability of LCA to identify the cases that the DSM-IV categories could not pick up. For instance, out of 502 cases of the “Predominantly Hyperactive–Impulsive” latent class, the DSM-IV Hyperactive Impulsive subtype only picked up 13 (2.59%) cases, whereas the “Predominantly Hyperactive–Impulsive” latent class identified a further 489 (97.4%) cases. In total, there were 1,194 (18.27%) cases that the DSM-IV ADHD diagnostic criteria could not pick up. The ability of LCA to pick up more cases than DSM-IV does not mean that all chosen individuals by LCA are clinical cases.
|LC-9 groups||Overlapping cases1||Distinct cases2||Total cases|
|Few symptoms||4,375 (98.94%)||47 (1.06%)||4,422 (100%)|
|Predominantly Hyp–Imp||13 (2.59%)||489 (97.41%)||502 (100%)|
|Moderate RD||238 (65.7%)||124 (34.3%)||362 (100%)|
|Predominantly Inattentive||28 (7.2%)||361 (92.8%)||389 (100%)|
|Severe RD||227 (97.4%)||6 (2.6%)||233 (100%)|
|Predominantly Inattentive & RD||169 (92.86%)||13 (7.14%)||182 (100%)|
|Combined||26 (14.44%)||154 (85.56%)||180 (100%)|
|Unique Severe RD||147 (100%)||0 (0.0%)||147 (100%)|
|Combined & RD||118 (100%)||0 (0.0%)||118 (100%)|
|Total||5,341 (81.73%)||1,194(18.27%)||6,535 (100%)|
- Top of page
- MATERIALS AND METHODS
This study applied LCA to ADHD and RD subtypes in order to identify genetic phenotypes that would be effective in performing a genotyping analysis. Nine latent classes were produced which could be divided into three groups; a group containing three ADHD latent classes, which matched the DSM-IV ADHD subtypes; a second group containing three RD latent classes; and a third group containing two ADHD–RD comorbid latent classes. The symptoms within each of these nine latent classes displayed greater homogeneity, and the existence of such clearly-delineated symptom clusters allows for greater etiological precision in identifying the influence of genetics on ADHD and its subtypes.
The prevalence of ADHD/RD latent subtypes showed that the most prevalent subtype was “Predominantly Hyperactive–Impulsive” (7.7%). However, by adding the prevalence of all Inattentive latent classes, including the “Predominantly Inattentive” (6.0%) and the “Predominantly Inattentive and Reading Disability” subtypes (2.8%), the total prevalence of the Predominantly Inattentive subtypes (8.2%) would be the highest among the latent classes. The total prevalence for both “Combined” and “Combined and Reading Disability” was 4.6%.
Rasmussen et al. [2002b, 2004] previously applied LCA to an ADHD Australian male and female sample, in order to replicate their findings with adolescent female twins from Missouri. Both studies found the eight latent class model was the best fit. The classes in this model included one “Few Symptom” class, three “mild to moderate ADHD” classes, three “severe ADHD” classes, and one “unique ADHD” class. However, the study considered only six latent classes. The 18 DSM-IV ADHD item ratings for two latent classes were different as they were associated with broad confidence intervals. Comparing the latent classes from Rasmussen et al.'s study [2002b] with the current study, both studies share the same number of ADHD latent classes, after excluding the two latent classes described above from Rasmussen et al.'s study [2002b]. There were slight phenotypic differences in relation to ADHD severity between the two studies.
An examination of gender differences among the ADHD–RD latent classes, excluding unaffected individuals, showed that there were more affected boys than girls: 20.1–12.3%. However, it was not obvious if there was a relationship between the structure of LC-9 and gender differences in the classes. Differences in gender prevalence occurred in each latent class. The “Predominantly Inattentive” latent class had slightly more girls (0.3%) than boys (0.1%), which could be due to the presence of higher genetic contributions between the domains in females than in males.
Rasmussen et al. [2002a] stated that when using LCA, individuals with comparable ratings across all symptoms, based on statistical probabilities, form natural groups. This conclusion does not take into account the total number of symptoms or the presence of impairment. Based on these findings, this idea can be further developed to compare the accuracy of both ADHD and RD continuous and categorical data. The ratings for the 18 items were found to be higher for the DSM-IV subtypes; however, LCA proved to be more effective in identifying the presence of ADHD in individuals overall. This conclusion is based on the results above, which show that LCA identified a greater number of cases of ADHD than the DSM-IV approach.
The seven RD item endorsements showed relatively consistent correspondence with the severity of the RD phenotype. The more highly the RD items are endorsed, the more severe is the RD phenotype. However, the RD item endorsements with the three RD latent classes showed a minor discrepancy. Rather than having strong RD item endorsement for “Slow reading more than other children of the same age,” and “Reading below expectancy level” with the “Unique Severe RD,” these items were endorsed more highly with the “Severe RD” latent class. The reason for this could be that the LCA clustering for these two items tended to have greater correspondence with the “Severe RD” latent class instead of the “Unique Severe RD” latent class.
The weakest RD item endorsements among the seven items were “Difficulty learning the days or months” and the “Difficulty learning letter names” items. This might be due to the weak validity and reliability found by Willcutt et al. [2003a] for the item “Difficulty learning letter names.” This could also be applicable to the “Difficulty learning letter names” item, as it also showed a minor cross-loading (0.34) on the factor analysis. Although these two items were endorsed strongly (0.90 and 0.87 respectively) with the “Unique Severe RD” latent class, their endorsement was extremely low with the “Moderate RD” (0.25 and 0.07) and “Severe RD” (0.17 and 0.10) latent classes. This indicates that the LCA might effectively cluster these two symptoms and assigned them to the “Unique Severe RD” latent class.
There were higher concordance rates among MZ twins than DZ twins, except for the “Few Symptoms” and the “Predominantly Inattentive” latent classes. The concordance among MZ and DZ pairs was explored to determine the presence of genetic effects that contribute to the ADHD–RD latent classes. As the ADHD–RD latent classes were significantly higher among MZ twins than DZ twins, this indicates a presence of common genetic effects for MZ twins. However, two latent classes where DZ twins rated slightly higher than MZ twins were the “Few Symptoms” and the “Predominantly Inattentive” classes.
Willcutt et al.'s [2003b] findings of the presence of genetic entities due to attributed genetic influence in the comorbidity between ADHD and RD for the “Predominantly Inattentive/RD” and the “Combined/RD” latent classes did not show the same pattern of zygosity in this study. The former latent class was nearly the same in the number of the concordant MZ and DZ twin pairs, indicating lower genetic contribution to this than to the latter latent class, in which the concordant MZ twin pairs were double the number of the concordant DZ twins, indicating a high genetic contribution. There were significant differences between the concordant MZ twins and concordant DZ twins, with MZ twin concordance being double that of DZ twins, indicating that the latent classes are heritable and may contain genes that contribute to the three ADHD latent classes.
This is the first time LCA has been used to assess the comorbidity of ADHD and RD. Effectively, LCA confirmed the ADHD–RD comorbidity by creating two distinctive latent classes; the “Predominantly Inattentive/RD” and “Combined/RD” latent classes. The former latent class replicated previous findings for the presence of a shared genetic contribution between the Inattentive subtype and RD [Willcutt et al., 2000, 2003b]. However, LCA did not produce a comorbid class for RD with the “Predominantly Hyperactive–Impulsive” latent class that was also supported by a low shared genetic contribution between the Hyperactive–Impulsive subtype and RD [Willcutt et al., 2000, 2003b]. This study asserts that both the “Predominant Inattentive RD” and the “Combined RD” latent classes are significant and can help in understanding the comorbidity between ADHD and RD. Each latent class has a homogenous and genetically informative phenotype helping to understand the comorbidity between the two domains. These two latent classes conform to the strong comorbid relationship of the DSM IV Inattentive and Combined subtypes with RD, and the weak comorbid relationship of the Hyperactive–Impulsive subtype with RD, which might have clinical implications that require more investigation.
Interestingly, LCA was also able to demonstrate that RD is strongly comorbid with the Combined subtype, by creating the “Combined/RD” latent classes. The Chi-square test for zygosity of the “Combined/RD” latent class showed significant differences among concordant and discordant MZ and DZ twins, indicating the presence of a higher genetic factor among this latent class. However, the Chi-square test for zygosity of the “Predominantly Inattentive RD” latent class was not significant among concordant and discordant MZ and DZ twins. The ability of LCA to pick up more cases than DSM-IV does not mean that all chosen individuals by LCA are clinical cases. Previously, Hudziak et al. , Neuman et al. , and Rasmussen et al. [2002b] used LCA to refine ADHD classification in population-based samples of children through combining probability of symptom endorsement and overall symptom profile to create a set of eight distinct classes of ADHD symptoms. These classes are Few Symptoms, Mild combined, Mild Inattentive, Severe Combined, Severe Inattentive, Moderate Inattentive, Severe Hyperactive–Impulsive, and Moderate Combined. Although five latent classes are not clinically relevant, three of these classes are so, which are Severe Combined, Severe Inattentive, and Severe Hyperactive–Impulsive. Todd et al.  explained that the severe inattentive latent class was associated with academic problems, family problems, and referral to health care providers.
In the current study, the age range of the sample was 4–18 years. Given that ADHD symptoms have been found to change with age [Barkley, 1997], and Todd et al.  found that ADHD phenotypes based on LCA were unstable over a 5-year period when derived from cross-sectional data, it is important to consider the influence of factors such as age and developmental course on phenotypes that may be used in genetic studies.
- Top of page
- MATERIALS AND METHODS
LCA was effective in refining the phenotypes of ADHD alone, RD alone, and ADHD–RD comorbidity, and classified these into homogenous groups based on clusters of symptoms. The only class that did not demonstrate heritability was the “Predominantly Inattentive/RD” latent class based on comparing the rates of MZ twins versus DZ twins. It is also suggested that the comorbid ADHD–RD latent classes may be genetically distinctive from ADHD alone and RD alone. The LCA's production of the nine ADHD–RD latent classes comes from its ability to dissect phenotypes into several robust dimensions.
Willcutt  explained that “because LCA approach often includes a substantially higher number of individuals for each class, this may increase power to detect differences between classes in comparison to analyses of the DSM-IV subtypes.” The performance of LCA in identifying cases was also significantly better in most of the other latent classes than was the DSM-IV defined categories. The phenotypes of LCA cases may be genetically informative for use in genotyping analysis. Although previous literature (e.g., Todd et al. ) has demonstrated that LCA may be a better approach in genotyping analysis, and may be more genetically informative than DSM-IV, further investigation is required as other factors, genetic and environmental, may influence symptom clusters.
The recent advances in the field of psychiatric genetics have moved the knowledge of phenotypes to a new era called phenomics (the systematic cataloguing of phenotypes on a genome-wide scale) [Bilder, 2008], which is consolidated by the amazing progress in bioinformatics. The latter is utilized to group genomic information according to assumed tasks in signalling pathways, as well as being applied as a rational extension for similar strategies on higher level phenotypic data [Bilder, 2008]. To reduce phenotypic complexity, bioinformatics strategies can narrow this problem down and help to reveal important relations between phenotypes [Sabb et al., 2009]. Accordingly, this study aims to contribute to an online (web-based) collaborative database for phenotype annotation called “Phenowiki.” PhenoWiki currently provides methods to document quantitative effects supporting the validity of phenotypic concepts, heritability of these phenotypes, and selected relations between phenotypes [Bilder et al., 2009].
Although the era of phenomics is still in its primitive stages of development [Bilder et al., 2008], it would be a promising approach for mapping human phenotypes on a genome-wide scale. Integrating both the traditional approach represented by mathematical models such as LCA with the rise of the new era of phonemics and bioinformatics can play an important role in validating phenotypic concepts, revealing the understanding of etiology and classification of psychiatric disorders [Bilder et al., 2009; Sabb et al., 2009]. Therefore, the produced sub-phenotypes might be validated into phenotypic concepts and put in the online collaborative database for phenotype annotation “Phenowiki,” in order to contribute in the building of a phenotypic catalogue for ADHD–RD comorbidity, and introduce a probable and a better understanding of relations between ADHD–RD phenotypes to be used for further investigations into comorbidity, thus increasing the understanding of etiology and classification of psychiatric disorders.
- Top of page
- MATERIALS AND METHODS
We would like to thank the National Health and Medical Research Council (NHMRC) (Grant ID: 229005) for funding this project. The authors also would like to thank the Australian Twin Registry (ATR) for providing us with their recruited Australian twin families nation-wide. We are grateful to all study participants who gave their time to complete our questionnaires. Also, sincere thanks go to Dr. Kellie Bennett for her co-operation and assistance in performing the Latent Class Analysis (LCA).
- Top of page
- MATERIALS AND METHODS
- American Psychiatric Association. 1980. Diagnostic and statistical manual of mental disorders. Washington, DC: American Psychiatric Association.
- American Psychiatric Association. 1987. Diagnostic and statistical manual of mental disorders. Washington, DC: American Psychiatric Association.
- American Psychiatric Association. 1994. Diagnostic and statistical manual of mental disorders. Washington, DC: American Psychiatric Association.
- American Psychiatric Association. 2000. Diagnostic and statistical manual of mental disorders text revision. Washington, DC: American Psychiatric Association.
- 1997. ADHD and the nature of self-control. New York: Guilford Press. .
- 2006. The Australian Twin ADHD Project: Current status and future directions. Twin Res Hum Genet 9: 718–726. , , , , , .
- 2005. Attention deficit hyperactivity disorder. Lancet 366(9481): 237–248. , .
- 2008. Phenomics: Building scaffolds for biological hypotheses in the post-genomic era. Biol Psychiatry 63: 439–440. .
- 2009. Phenomics: The systematyic study of phenotypes on a genome-wide scale. Neuroscince 146(1): 30–42. , , , , , , , , .
- 2012. Aetiology for the covariation between combined.type ADHD and reading difficulties in a family study: The role of IQ. JChild Psychol Psychiatry DOI: 10.1111/j.1469-7610.2012.02527.x , , , , , , , , , , et al.
- 1975. Reliably separating identical from fraternal twins. Arch Gen Psychiatry 32: 1371–1375. , , , .
- 1993. An epidemiological study of disorders in late childhood and adolescence—I. Age- and gender-specific prevalence. J Child Psychol Psychiatry 34: 851–867. , , , , , , , , .
- 1977. Maximum likelihood from incomplete data via the EM algorithm. J R Stat Soc Ser B 39: 1–38. , , .
- 2007. Structure of the Autism symptom phenotype: A proposed multidimensional model. Am Acad Child Adolesc Psychiatry 46(2): 188–196. , , , , , , , .
- 2004. Twin and Sibling Questionnaire. Unpublished Questionnaire, Curtin University of Technology, Perth, WA. , .
- 2001. Introduction to the genetic analysis of attentional disorders. In: Levy F, Hay DA, editors. Attention, genes and ADHD. Philadelphia: Brunner-Routledge. pp 7–34. , , .
- 2002. Recruitment and attrition in twin registrar studies of childhood behavior: The example of the Australian Twin ADHD Project. Twin Res 5(5): 324–328. , , , .
- 2003. Zygosity diagnosis in the absence of genotypic data: An approach using latent class analysis. Twin Res 6(1): 22–26. , , , , , .
- 1998. Latent class and factor analysis of DSM-IV ADHD: A twin study of female adolescents. J Am Acad Child AdolescPsychiatry 37: 848–857. , , , , , , , , .
- 2006. The genetics of ADHD: A literature review of 2005. Curr Psychiatry Rep 8: 393–397. , .
- 2001. Attention genes and ADHD. Philadelphia: PN Brunner-Routledge. p. 263. , .
- 1996. Twin-sibling differences in perinatal reports of ADHD, speech, reading and behavior problems. JChild Psychol Psychiatry 37: 569–578. , , , , .
- 1997. Attention-deficit hyperactivity disorder: A category or a continuum? Genetic analysis of a large-scale twin study. J Am Acad Child Adolesc Psychiatry 36: 737–744. , , , .
- 2005. Gender differences in ADHD subtype comorbidity J Am Acad Child Adolesc Psychiatry 44(4): 368–376. , , , .
- 1987. Latent class analysis. Newbury Park, CA: Sage Publications. .
- 2006. Breakthroughs in the search for dyslexia candidate genes. Trends Mol Med 12(7): 333–341. , , .
- 2011. A dyslexia-associated variant in DCDC2 changes gene expression. Behav Genet 41: 58–66. , , , , , , , , .
- 1999. Evaluation of ADHD typology in three contrasting samples: A latent class approach. J Am AcadChild Adolesc Psychiatry 38(1): 1–14. , , , , , , , , , , et al.
- 2005. Estimation of prevalence of DSM-IV and latent-defined ADHD subtypes in a population-based sample of child and adolescent twins Twin Res Hum Genet 8(4): 392–401. , , , , , , .
- 1966. The diagnosis of twin zygosity. Acta Genet Stat Med 16: 265–275. , .
- 2006. From single to multiple deficit models of developmental disorders. Cognition 101: 385–413. .
- 2009. Diagnosing learning disorders: A neuropsychological framework, 2nd edition. New York, NY: The Guilford Press. .
- 2002a. Comparison of male adolescent-report of attention-deficit/hyperactivity disorder (ADHD) symptoms across two cultures using latent class and principal components analysis. J Child Psychol Psychiatry 43(6): 797–805. , , , , , .
- 2002b. Replication of the latent class structure of Attention-Deficit/Hyperactivity Disorder (ADHD) subtypes in a sample of Australian twins. J Child Psychol Psychiatry 43(8): 1018–1028. , , , , , .
- 2004. Familial clustering of latent class and DSM-IV defined attention-deficit/hyperactivity disorder (ADHD) subtypes. J Child Psychol Psychiatry 45(3): 589–598. , , , , , .
- 2009. Challenges in phenotype definition in the whole-genome era: Multivariate models of memory and intelligence. Neuroscince 164: 88–107. , , , , , , , , , , et al.
- 1990. Prevalence of reading disability in boys and girls. Results of the Connecticut longitudinal study. J Am Med Assoc 264: 998–1002. , , , .
- 2007. Attention-deficit/hyperactivity disorder: A neuropsychological perspective towards DSM-V. Neuropsychol Rev 17(1): 5–38. , .
- 2007. Informative phenotypes for genetic studies of psychiatric disorders. Am J Med Genet Part B 144B: 581–588. , , , , , , .
- 2007. Can comparison of MZ- and DZ-twin concordance rates be used invariably to estimate heritability? Twin Res Hum Genet 10(5): 712–717. , .
- 2001. Familiality and heritability of subtypes of attention deficit hyperactivity disorder in a population sample of adolescent female twins. Am J Psychiatry 158: 1891–1898. , , , , , , , .
- 2002. Discrimination of DSM-IV and latent class attention-deficit/hyperactivity disorder subtypes by educational and cognitive performance in a population-based sample of child and adolescent twins. J Am Acad Child Adolesc Psychiatry 41(7): 820–828. , , , , , .
- 2005. Collaborative analysis of DRD4 and DAT genotypes in population-defined ADHD subtypes. J Child Psychol Psychiatry 46(10): 1067–1073. , , , , , , , , .
- 2008. Predictors of stability of attention-deficit/hyperactivity disorder subtypes from childhood to young adulthood. J Am Acad Child Adolesc Psychiatry 47(1): 76–85. , , , , , , .
- 2006. Validation of population-based ADHD subtypes and identification of three clinically impaired subtypes. Am J Med Genet Part B 141B: 312–318. , , , .
- 2008. In: Sheikhi A, editor. Thesis examination. Perth. .
- 2000. Twin study of the etiology of comorbidity between reading disability and attention-deficit/hyperactivity disorder. Am J Med Genet Part B 96: 293–301. , , .
- 2003a. Internal and external validty ratings of reading difficulties. Boulder. , , , .
- 2003b. Genetic etiology of comorbid reading difficulties and ADHD. In: Plomin R, DeFries J, McGuffin P, Craig I, editors. Behavioral genetics in a postgenomic era. Washington DC: American Psychological Association. pp. 227–246. , , , , , .
- 2011. Colorado learning difficulties questionnaire: Validation of a parent-report screening measure for learning difficulties. Psychol Assess 23(3): 778–791. , , , , , .
- 2010. Rethinking shared environment as a source of variance underlying attention-deficit/hyperactivity disorder symptoms: Comment on Burt (2009) Psychol Bull 136: 331–340. , , , , .
- 2007. Focus on Words: A twin study of reading and inattention. Behav Genet 37: 284–293. , , .