Fig. 8.1
Rates of comorbid psychopathology in groups RD only, ADHD only, RD + ADHD, and neither disorder. Bars with different letters are significantly different (P < 0.05)
Clinical implications. Studies of the external correlates of developmental psychopathologies further underscore the clinical importance of comorbidity. Studies of a number of different disorders suggest that individuals with at least one comorbid diagnosis are often more impaired than individuals who do not meet criteria for any other disorders (e.g., Connor, Steeber, & McBurnett, 2010; Goldstein & Levitt, 2008; Waschbusch, 2002; Willcutt & McQueen, 2010). This pattern is clearly illustrated by our findings in the CLDRC sample, which indicated that in comparison to groups with RD or ADHD alone, individuals with both RD and ADHD exhibited greater academic, social, and neuropsychological impairment, were more likely to be retained or expelled from school, and were more likely to have had contact with the juvenile justice system (Willcutt et al., 2001; Willcutt, Betjemann, Pennington, et al., 2007; Willcutt, Pennington, et al., 2005).
Overview of the Chapter
The overall goal of the current chapter is to provide an overview of behavioral genetic methods that can be used to test competing explanations for comorbidity between complex psychopathologies. The first section of the body of the chapter summarizes the most prominent competing theoretical explanations for comorbidity, and subsequent sections then describe behavioral genetic methods that can be used to directly test these competing models. Because space constraints preclude a comprehensive review of studies of all possible pairs of comorbid disorders, the methods are primarily illustrated by examples from our work examining the etiology of comorbidity between RD, ADHD, and other disorders (for a detailed summary of these studies, see Willcutt, Pennington, et al., 2010). Finally, the chapter concludes by highlighting several directions for future research that are likely to facilitate additional refinements to diagnostic models of developmental psychopathology by providing additional insight regarding the etiology of comorbidity.
Current Issues
Theoretical Models of Comorbidity
Artifactual Comorbidity
Before conducting a behavior genetic study to examine the etiology of comorbidity between disorders, it is important to first test whether the observed comorbidity is simply an artifact caused by a biased sampling procedure or measurement problem (e.g., Angold et al., 1999). Artifactual comorbidity could occur due to ascertainment biases in clinic-referred samples, rater biases or shared method variance if the same rater is used to define both disorders, or symptom overlap between disorders. Data from our group and others suggest that these artifactual hypotheses can be clearly rejected as explanations for comorbidity between RD and ADHD and most other pairs of disorders (e.g., Angold et al., 1999; Willcutt, Pennington, et al., 2010).
Competing Explanations of Non-artifactual Comorbidity
Over a dozen competing models have been proposed to account for non-artifactual comorbidity between developmental psychopathologies (for a comprehensive discussion, see Neale & Kendler, 1995; Rhee, Hewitt, Corley, Willcutt, & Pennington, 2005). Because space constraints preclude a detailed discussion of all of these models, the remainder of this section focuses on four of the most prominent competing explanations for comorbidity between RD, ADHD, and a range of internalizing and externalizing disorders.
The correlated liabilities model suggests that two disorders co-occur more often than expected by chance due to shared etiological influences, whereas the disorders are distinguished by other etiological influences that are specific to each disorder. As described throughout the remainder of this chapter, behavioral genetic approaches are especially well-suited to test the correlated liabilities model.
Causal models suggest that one disorder directly causes the underlying pathophysiology weaknesses and behavioral manifestation of the second disorder. For example, if severe ADHD symptoms interfere with a child’s ability to attend to early classroom instruction on the specific phonological processing skills that underlie the development of reading, these attentional difficulties could directly cause weaknesses in both phonological processing and reading development. Behavioral genetic methods can also test different causal models, although many of these approaches require access to etiologically informative longitudinal data.
The three independent disorders model suggests that the comorbid group is a third disorder with an etiology that is distinct from the causal factors that lead to each disorder alone. This model would be supported if family studies revealed that the three groups were transmitted independently or if twin studies reported that the genetic and environmental influences on one of the disorders varied significantly as a function of comorbidity with the other disorder.
Finally, phenocopy models (also called multiformity models) suggest that one disorder causes the phenotypic manifestation of the second disorder in the absence of the underlying pathophysiological dysfunction that increases liability to the second disorder when it occurs in isolation. For example, a child with severe reading difficulties might appear inattentive or off task during reading assignments in the classroom due to their inability to read rather than any specific attentional difficulties. Behavior genetic studies of behavioral symptoms often have difficulty distinguishing between the phenocopy and correlated liability models because both predict similar behavioral outcomes. To address this limitation, recent behavioral genetic studies have begun to also incorporate measures of the specific neuropsychological weaknesses that characterize each disorder. An example of this approach is described in more detail in the final section of the chapter.
Behavior Genetic Approaches to Understand Comorbidity
Family and twin studies provide critical tests of many of the competing models that have been proposed to account for comorbidity between multifactorial disorders. In this section I first describe univariate analyses to illustrate the rationale for each approach and provide a framework for interpretation of subsequent analyses of the etiology of comorbidity. Examples from the CLDRC or other projects are then used to illustrate the implementation of each method and the subsequent interpretation of results.
Univariate Family Studies
Family studies test whether a trait or disorder is correlated among biological relatives in a family. If genetic or family environmental influences increase susceptibility to a disorder, the disorder should occur more frequently in relatives of individuals with the disorder than in relatives of individuals without the disorder. Studies of psychopathology have tested for familiality by examining correlations between family members on dimensional measures of symptoms, along with concordance rates for categorical diagnoses.
Dimensional analyses. Previous results from the CLDRC and other family studies indicate that correlations between biological siblings are significant and medium to large in magnitude for dimensional measures of reading (r = 0.40–0.70) and ADHD symptoms (r = 0.20–0.50; e.g., Byrne et al., 2007; McLoughlin, Ronald, Kuntsi, Asherson, & Plomin, 2007; Petrill et al., 2007; Willcutt, Betjemann, Wadsworth, et al., 2007). The column in Table 8.1 labeled “sibling correlations within disorder” summarizes new analyses of the CLDRC sample that were conducted to estimate the familiality of symptom dimensions that typically covary with measures of reading and ADHD symptoms (only DZ twins and sibling pairs were included in these analyses to obtain an estimate of familiality that is not biased by the inclusion of identical twins). Sibling correlations were significant for all phenotypes that were included in these analyses, providing initial support for the familiality of each of these disorders.
Table 8.1
Phenotypic correlations and within-trait and cross-trait sibling correlations between dimensional measures of ADHD symptoms, reading, and comorbid psychopathology
Phenotypic correlations | Sibling correlations | ||||||
|---|---|---|---|---|---|---|---|
Within individuals | Between disorders | ||||||
Reading | Inattention | Hyp-imp | Within Disorder | Reading | Inattention | Hyp-imp | |
DBRS ADHD symptoms | |||||||
DSM-IV inattention | 0.35a | – | 0.60b | 0.31 | 0.19a | – | 0.29b |
DSM-IV hyperactivity-impulsivity | 0.16a | 0.60b | – | 0.29 | 0.06a | 0.29b | – |
Academic achievement | |||||||
PIAT reading | – | 0.35a | 0.16b | 0.43 | – | 0.19a | 0.06b |
PIAT math | 0.60a | 0.32b | 0.17c | 0.45 | 0.44a | 0.16b | 0.05b |
Externalizing symptoms | |||||||
DICA-IV oppositional defiant disorder | 0.16a | 0.36b | 0.51c | 0.40 | 0.07a | 0.17a | 0.29b |
DICA-IV conduct disorder | 0.14a | 0.33b | 0.33b | 0.45 | 0.09 | 0.20 | 0.17 |
CBCL/TRF delinquent behavior | 0.21a | 0.41b | 0.40b | 0.38 | 0.13a | 0.26b | 0.28b |
CBCL/TRF aggressive behavior | 0.18a | 0.41b | 0.58c | 0.43 | 0.09a | 0.22b | 0.34c |
APSD callous/unemotional scale | 0.19a | 0.47b | 0.37c | 0.24 | 0.09 | 0.14 | 0.23 |
APSD narcissistic scale | 0.14a | 0.37b | 0.51c | 0.21 | 0.07 | 0.18 | 0.24 |
Internalizing symptoms | |||||||
DICA-IV generalized anxiety disorder | 0.14 | 0.16 | 0.15 | 0.48 | 0.09 | 0.10 | 0.08 |
DICA-IV major depressive disorder | 0.14a | 0.35b | 0.19a | 0.33 | 0.08 | 0.18 | 0.10 |
CBCL/TRF withdrawn | 0.15a | 0.36b | 0.11a | 0.23 | 0.10 | 0.17 | 0.06 |
Concordance rates in relatives. Previous analyses by our group and others indicate that the relative risk for a diagnosis of RD is 4–8 times higher in first-degree relatives of probands with RD than in relatives of individuals without RD, and similar familiality has been reported for ADHD (e.g., DeFries, Singer, Foch, & Lewitter, 1978; Faraone, Biederman, & Friedman, 2000; Friedman, Chhabildas, Budhiraja, Willcutt, & Pennington, 2003). Similarly, the white bars in Fig. 8.2 indicate significant familial risk for categorical diagnoses of math disability (MD), oppositional defiant disorder (ODD), conduct disorder (CD), generalized anxiety disorder (GAD), and major depressive disorder (MDD), a pattern that is highly consistent with results reported by previous family studies (e.g., Burt, McGue, Krueger, & Iacono, 2005; Czajkowski, Roysamb, Reichborn-Kjennerud, & Tambs, 2010; Gross-Tsur, Manor, & Shalev, 1996; Rice, Harold, & Thapar, 2002).
Fig. 8.2
Concordance between siblings for math disability, ODD, CD, MDD, and GAD, and cross-concordance between siblings for RD and ADHD and each of these disorders. Relative risk in comparison to siblings of control probands was significant for all comparisons (P < 0.05)
Family Studies of Comorbidity
If two comorbid disorders are each significantly familial, a straightforward extension of univariate family-based analyses can be used to test whether shared familial risk factors contribute to comorbidity between the disorders. Rather than comparing the scores of two family members on a measure of a single disorder, family analyses of comorbidity examine the association between one individual’s score on the measure of the first disorder and their relative’s score on the measure of the second disorder. Significant cross-disorder covariance or concordance between relatives indicates that comorbidity between the two disorders is due at least in part to shared familial risk factors. To illustrate this approach, the CLDRC sample was used to examine the association between RD or ADHD scores in one twin and symptoms of learning, internalizing, and externalizing disorders in their siblings.
Cross–trait sibling correlations. Consistent with the significant phenotypic correlations between the dimensional measures of ADHD and RD and all measures of math and internalizing and externalizing symptoms, most cross-trait sibling correlations were significant (final three columns of Table 8.1). However, many of these correlations were small in magnitude, and cross-trait sibling correlations were not significant between hyperactivity-impulsivity and academic achievement or reading and some aspects of externalizing psychopathology.
Cross–concordance analyses. Although not all studies reported evidence of shared familial influences on learning disabilities and ADHD (e.g., Doyle, Faraone, DuPre, & Biederman, 2001), previous results from the CLDRC suggested that family members of probands with RD or ADHD were approximately three times more likely to meet criteria for the other disorder than family members of comparison probands without either disorder (e.g., Friedman et al., 2003). Similar analyses were conducted to test for co-familiality between RD and ADHD and other learning, externalizing, and internalizing disorders (the gray bars in Fig. 8.2 indicate co-familiality with ADHD and the black bars represent co-familiality with RD). In comparison to siblings of probands without RD, siblings of probands with RD were seven times more likely to meet criteria for math disability and over twice as likely to meet criteria for each internalizing and externalizing disorder. Similarly, siblings of probands with ADHD were approximately four times more likely than siblings of control probands to meet criteria for math disability, ODD, or CD, and more than twice as likely to meet criteria for MDD or GAD.
In addition to providing initial information regarding shared familial risk factors, family data also provide a key test of the three disorders model. This model makes the strong prediction that the familial risk factors that lead to comorbidity are distinct from the familial influences associated with each disorder when it occurs alone. Therefore, the three disorders model would be supported if family members of comorbid probands exhibited higher rates of comorbidity, but did not exhibit higher rates of either disorder alone. Contrary to this hypothesis, siblings of probands with comorbid RD + ADHD exhibited higher rates of RD alone (29 %), ADHD alone (18 %), and RD + ADHD (15 %) than siblings of comparison probands (all less than 5 %), suggesting that RD and ADHD are related disorders that sometimes co-occur due to shared familial risk factors.
Conclusions from family studies. Results of family studies clearly suggest that shared familial influences contribute to comorbidity between RD, ADHD, and a range of other psychopathology. In contrast, these data argue strongly against the three disorders model that suggested that the comorbid group is a third disorder that is unrelated to either disorder when they occur alone. The significant shared familial influences may potentially reflect common genetic influences; however, family studies cannot provide conclusive evidence regarding genetic effects because members of biological families living in the same home share both genetic and family environmental influences. Instead, other behavioral genetic designs such as twin studies are necessary to disentangle the relative contributions of genetic and environmental influences.
Twin Studies
By comparing the similarity of monozygotic (MZ) twins, who share all of their genes, to dizygotic (DZ) twins, who share half of their segregating genes on average, twin analyses provide estimates of the extent to which individual differences or extreme scores are due to genetic or environmental influences (e.g., Plomin, DeFries, Knopik, & Neiderhiser, 2013). Heritability is defined as the proportion of the total phenotypic variance in a trait that is attributable to genetic influences. The proportion of variance due to environmental factors is subdivided to distinguish two types of environmental influences. Shared environmental influences are environmental factors that increase the similarity of individuals within a family in comparison to unrelated individuals in the population. These effects may potentially include environmental influences within the home or any other shared experiences such as mutual friends or shared teachers. In contrast, nonshared environmental influences specifically affect only one member of a twin pair or are perceived differently by the two twins. Nonshared environmental risk factors could include a head injury or other accident, a traumatic event, or exposure to physical or sexual abuse (if the other twin was not similarly exposed).
Individual Differences
Univariate analyses. We recently reviewed all published twin studies of individual differences in reading, math, and symptoms of ADHD (Willcutt, Pennington, et al., 2010). Because a recent meta-analysis of DSM-IV ADHD strongly supported the validity of the distinction between two correlated but distinct dimensions of inattention and hyperactivity-impulsivity symptoms (Willcutt et al., 2012), these dimensions were examined separately.
Overall, heritability estimates based on variance components analyses of unselected samples of twins are moderate to high for individual differences in single-word reading and mathematics (heritability = 0.40–0.80) and high for symptoms of inattention and hyperactivity-impulsivity. Based on twin studies, shared environmental influences account for an additional 10–15 % of the variance in reading, but play a minimal role in the etiology of ADHD symptoms. Nonshared environmental influences and measurement error typically explain 10–20 % of the variance in each of the phenotypes.
Analyses of covariance between dimensions of psychopathology. Studies of several large twin samples have examined the etiology of covariance between symptoms of ADHD and individual differences in reading (Ebejer et al., 2010; Greven, Harlaar, Dale, & Plomin, 2011; Greven, Rijsdisjk, Asherson, & Plomin, 2012; Hart et al., 2010; Hay, Bennett, Levy, Sergeant, & Swanson, 2007; Paloyelis, Rijsdijk, Wood, Asherson, & Kuntsi, 2010; Sheikhi, Piek, Martin, & Hay, 2008; Willcutt, Betjemann, et al., 2010; Willcutt, Betjemann, Wadsworth, et al., 2007). Most of these studies indicate that common genetic influences contribute to the covariance between individual differences in reading and both DSM-IV ADHD symptom dimensions, but shared genetic influences are stronger for inattention than hyperactivity-impulsivity.
As reviewed in detail in the chapter by Wood, Rommel, and Kuntsi (2014, Chap. 1), analyses of unselected twin samples have also been used to examine the etiology of covariance between ADHD symptoms and antisocial behavior and features of autism spectrum disorders. In contrast, few studies of unselected samples have tested the etiology of covariance between ADHD and internalizing symptoms or between reading and any psychopathology, suggesting that these are important directions for future research.
Twin Studies of Selected Samples
Concordance and cross–concordance rates. The most straightforward test for genetic influences on a clinical disorder is the comparison of the rate of concordance in pairs of MZ versus DZ twins. If a disorder is influenced by genes, the proportion of pairs in which both twins meet criteria for the disorder should be higher in MZ pairs than DZ pairs. Results from the CLDRC indicate that the probandwise concordance rate is significantly higher in MZ twin pairs than DZ twin pairs for RD, ADHD, and each of the comorbid disorders measured in our sample (Table 8.2).
Table 8.2
Univariate concordance rates in MZ and DZ twin pairs for each disorder included in analyses of comorbidity
MZ pairs | Same-sex DZ pairs | Odds ratio | P | |||
|---|---|---|---|---|---|---|
N | Concordance (%) | N | Concordance (%) | |||
Learning disorders | ||||||
Reading disability | 147 | 74 | 131 | 44 | 3.6 | 5.8 × 10−7 |
Math disability | 135 | 65 | 116 | 46 | 2.2 | 0.003 |
Externalizing disorders | ||||||
ADHD | 131 | 72 | 118 | 27 | 7.0 | <1 × 10−10 |
ODD | 122 | 67 | 108 | 47 | 2.3 | 0.002 |
CD | 114 | 64 | 104 | 41 | 2.6 | 0.0007 |
Internalizing disorders | ||||||
GAD | 89 | 54 | 77 | 36 | 2.1 | 0.02 |
MDD | 78 | 49 | 68 | 29 | 2.4 | 0.01 |
A bivariate generalization of the univariate comparison of concordance rates in MZ and DZ pairs provides an initial test of the etiology of comorbidity between categorical disorders. Rather than comparing the concordance of MZ and DZ twins for the same disorder, the bivariate model examines how often the cotwin of a proband selected for one disorder meets criteria for the other disorder. When probands were selected for ADHD, cross-concordance rates with all other disorders were significantly higher in MZ pairs than DZ pairs (Table 8.3), suggesting that common genetic influences contribute to each of these comorbidities. Cross-concordance rates between RD and the other disorders were also higher in MZ pairs than DZ pairs, although the difference between MZ and DZ pairs was only marginally significant for GAD and MDD.
Table 8.3
Cross-concordance with comorbid disorders in twin pairs selected due to a proband with ADHD or RD
Cross-concordance in the cotwin | Proband selection | |||||
|---|---|---|---|---|---|---|
ADHDa | RDb | |||||
MZ (%) | DZ (%) | Oddsratio | MZ (%) | DZ (%) | Oddsratio | |
Learning disorders | ||||||
Reading disability | 38 | 22 | 2.2** | – | – | – |
Math disability | 35 | 20 | 2.2** | 43 | 29 | 1.8* |
Externalizing disorders | ||||||
ADHD | – | – | – | 37 | 18 | 2.7** |
ODD | 42 | 28 | 1.9* | 28 | 19 | 1.7* |
CD | 26 | 16 | 1.8* | 15 | 9 | 1.8* |
Internalizing disorders | ||||||
GAD | 27 | 16 | 1.9* | 19 | 14 | 1.4 |
MDD | 17 | 6 | 3.2** | 18 | 12 | 1.6 |
Multiple regression analysis. Although the simplicity of a comparison of concordance rates is appealing, increasing evidence suggests that RD, ADHD, and other psychopathologies are defined by a diagnostic threshold imposed upon an underlying quantitative distribution of liability that is continuously distributed in the population (e.g., Hudziak, Achenbach, Althoff, & Pine, 2007; Willcutt et al., 2012; Willcutt, Pennington, & DeFries, 2000a). If the dimensional model is indeed correct, the transformation of continuous measures of individual differences into a categorical diagnosis results in the loss of important information regarding both severity differences within the disorder and variability in subthreshold symptomatology. To address this concern, DeFries and Fulker developed a more powerful multiple regression approach to test the etiology of extreme scores when probands are selected due to high scores on a phenotype that is continuously distributed in the population (DF analysis; DeFries & Fulker, 1985, 1988).
DF analysis is based on the differential regression of MZ and DZ cotwin scores toward the population mean when probands are selected due to an extreme score. Although scores of both MZ and DZ cotwins are expected to regress toward the population mean, scores of DZ cotwins should regress further than scores of MZ cotwins to the extent that extreme scores are influenced by genes. After appropriate standardization and transformation of scores, the magnitude of differential regression by zygosity provides a direct estimate of the extent to which genetic influences (h 2 g) and shared environmental influences (c 2 g) contribute to the group deficit.
Related posts:
Sleep and Psychopathology: Quantitative and Molecular Genetic Research on Comorbidity
Genetic Influences on Depression and Anxiety in Childhood and Adolescence
Progress in Understanding the Causes of Autism Spectrum Disorders and Autistic Traits: Twin Studies from 1977 to the Present Day
Genetics of Substance Use Disorders
Additional Evidence for Meaningful Etiological Distinctions Within the Broader Construct of Antisocial Behavior
Gene–Environment Interactions (G×E) in Behavioral Genetics
Stay updated, free articles. Join our Telegram channel
Full access? Get Clinical Tree
