Methodological issues in behavioural phenotypes

The aim of behavioural phenotype research is to determine whether there is a causal link between a genetic syndrome and behaviour. In order to determine such a link, a number of methodological issues need to taken into account. These issues include statistical, recruitment and measurement considerations.

3.5.1 What to measure?

There are many potential lines of enquiry open to researchers in the field of behavioural phenotypes including the assessment of general behaviour disorder, psychiatric and developmental disorders, general and specific cognitive profiles (O’Brien, & Yule, 1995a). The most popular and often first line of enquiry has been to investigate the presence of general behavioural disorder in individuals with a genetic syndrome. There have been numerous studies describing the prevalence of general behaviour disorder in many different syndromes including Williams syndrome (Einfeld, Tonge, & Florio, 1997), Prader-Willi syndrome (Clarke, Boer, Chung, Sturmey, & et al, 1996; van Leishout, de Meyer, Curfs, Koot, & Fryns, 1998), Fragile-X syndrome (Baumgardner, Reiss, Freund, & Abrams, 1995; Einfeld, Tonge, & Florio, 1994), Smith-Magenis Syndrome (Dykens, Finucane, & Gayley, 1997), and Angleman syndrome (Summers & Feldman, 1999).

Profiling the pattern of general behaviour disorder may lead to other lines of enquiry such as the presence of psychiatric disorder in individuals with genetic syndromes. For example, a number of studies have investigated the

prevalence of schizophrenia in individuals with Velo-cardio-facial syndrome (e.g., Frydman, 1999; Murphy, Jones, & Owen, 1999) and the diagnosis of dementia and Alzheimer’s disease has been studied in individuals with Down’s syndrome (e.g.. Deb & Braganza, 1999; Holland, Hon, Hupped, Stevens, & Watson, 1998). Along a similar line, the presence of developmental disorders has also been investigated in individuals with genetic syndromes. For example, studies have investigated the presence of autistic disorder in individuals with Fragile-X syndrome (e.g. Einfeld, Molony, & Hall, 1989; Turk & Graham, 1997).

Another line of enquiry, which has often been adopted, is to identify specific cognitive profiles associated with genetic disorders. A number of studies have determined that individuals with different genetic disorders show particular uneven cognitive profiles. For example, individuals with Williams syndrome have been shown to have relatively good verbal abilities in comparison to deficits in visuospatial abilities (e.g. Udwin & Yule, 1991). This uneven cognitive profile has led others to investigate the cognitive processes adopted by individuals with Williams syndrome. For example, Deruelle, Mancini, Livet, Cassé-Perrot, and de Schonen (1999) investigated how individuals with Williams syndrome process faces. They found individuals with Williams syndrome show an abnormal developmental course of face processing.

These lines of enquiry are not limited to describing between-syndrome differences. As Dykens (1995) noted, not all individuals with a syndrome will show a characteristic of a syndrome. Another line of enquiry involves the investigation into factors contributing to the within-syndrome variability. For example, Skuse et al. (1997) noted within syndrome variability in females with Turner’s syndrome on a measure of social cognition. They found those individuals with a maternally derived X-chromosome had inferior verbal and executive skills compared to those with a paternally derived X-chromosome which mediated deficits in social interaction. Thus, the within syndrome variability in social cognition was related to genetic influences. Factors

involved in within-syndrome variability are likely to be numerous and complex but by measuring key variables (such as age and IQ) it may be possible to identify some of factors which play a role in any variability observed.

In summary, behavioural, psychiatric and developmental disorders could all be associated with genetic disorders and could form the focus of a behavioural phenotype investigation. In addition, there may be a particular associated cognitive profile which requires investigation (e.g., impaired or preserved social, visuospatial or linguistic abilities). Other variables, which may contribute to within-syndrome variability in the behavioural phenotype, could also be assessed.

3.5.2 Use of control and comparison groups and statistical tests

The first stage in determining a causal link between a syndrome and behaviour requires providing evidence for an association between the genetic syndrome and the behaviour in question. This evidence needs to be provided by at least a case-controlled study; a study which compares the rate at which a particular behaviour occurs in cases of the syndrome and an appropriate control group (Einfeld & Hall, 1994). The aim of the control group is to rule out spurious associations between the syndrome and behaviour due to third variables. Three main control designs have been employed in behavioural phenotype research.

First, the behaviour observed in individuals with a genetic syndrome could compared to that seen in the general population (e.g. Vicari, Carlesimo, Brizzolara, & Pezzini, 1996). This choice of control group is not appropriate for most behavioural phenotype research. Any difference in the behaviour seen in the two groups could be explained by many factors other than the presence of the genetic disorder including overall level of intelligence, specific cognitive abilities, and environmental influences. These factors are third variables which need to be ruled out as potential causes of the observed behaviour in the individuals with a genetic syndrome.

Many genetic syndromes are associated with some degree of intellectual impairment. It is a well-known fact that the prevalence of behavioural disturbance increases with level of intellectual impairment. For example, in a large epidemiological study Rutter, Graham, and Yule (1970) found that children with learning disabilities had three to four times more behaviour problems that children with an IQ in the normal range. This finding has since been confirmed by many different researchers (e.g., Campbell & Malone, 1991; Corbett, 1979; Einfeld & Tonge, 1996a; Einfeld & Tonge, 1996b; Gillberg et al., 1986). Therefore, level of intellectual impairment is one of the most important variables to rule out as a potential cause of the behavioural features seen in individuals with genetic syndromes associated with learning disabilities. There may also be different environmental influences acting on individuals with learning disabilities compared to normally developing individuals because of their different life experiences and difficulties (O’Brien & Yule, 1995b).

The second control group design used has been to compare individuals with a genetic syndrome to a group of individuals with intellectual impairments of mixed aetiologies. For example, Einfeld et al. (1994) assessed behavioural and emotional disturbance in 48 individuals with Fragile X syndrome. There were compared to 454 individuals with learning disabilities of mixed aetiologies matched for IQ, sex and age with the Fragile X group. The limitation of this approach is that the mixed group of individuals will contain those with genetic syndromes which may have their own associated behavioural phenotypes. Also, those currently without a diagnosis may have genetic syndromes not yet identified and diagnosed. The strength of such an approach is that individuals with the genetic syndrome are compared to individuals with a similar level of learning disability in general.

The third control group design used has been to compare those with the syndrome of interest to those with a different genetic syndrome. For example, in order to investigate the relationship between autism and Fragile- X, Turk and Graham (1997) compared 56 boys with Fragile X syndrome to

45 boys with Down’s syndrome. This approach has its limitations in that data are obtained on how one genetic syndrome differs from another. In order to profile a behavioural phenotype of a particular syndrome, comparisons may need to be made to many different syndromes making the number of potential control groups limitless. It seems a combination of approaches may be warranted to fully characterise a behavioural phenotype.

As the first stage in describing the behaviours associated with a genetic syndrome, a group of individuals with the syndrome could be compared to a group of individuals with the same level of intellectual impairment of mixed aetiologies. In this way, the behaviours that are related to the genetic syndrome rather than the level of intellectual impairment can be determined. Research can then progress to compare behaviours associated with the genetic syndrome in question to those seen in other genetic syndromes in order to investigate partial specificity effects, as suggested by Hoddap (1997). In fact, Turk and Graham (1997) used this approach. In addition to comparing their Fragile X group to individuals with Down’s syndrome, they also compared to boys with no known cause for their learning disabilities, matched to the Fragile X group on age and IQ.

As stated above, the selection of the control group depends on the potential confounding variables which need to be ruled out as explanations for the association between the syndrome and behaviour. Level of intellectual ability is not the only potential confounding variable. Age is also likely to have an impact upon behavioural features and may need to be controlled for in a behavioural phenotype study. The confounding variables may be different for different disorders. For example, due to the nature of the genetic mechanism, males with Fragile-X syndrome are more severely affected than females. Therefore, in any study profiling the behavioural phenotype of Fragile-X it will be important to control for the effect of gender.

There are two ways in which the control group can be used to rule out potential confounding variables. First, the control group can be chosen to

match the index group on these variables. Second, these confounding variables can the measured in both groups and their effects can be controlled for statistically.

To test the association between syndromes and behaviour, it is necessary to employ statistical techniques for two reasons (Einfeld & Hall, 1994). First, by using an appropriate test of statistical significance, it can be shown that the association between the syndrome and behaviour is unlikely to have occurred by chance. Second, the use of statistical techniques also allows for the evaluation of the possibility that an effect was present but missed (Type II error) by conducting post-hoc power calculations.

In summary, each genetic disorder may have associated characteristics which need to be ruled out as potential causes of the behavioural manifestations (e.g., level of learning disability, age). For each genetic disorder, it is necessary to consider which variables to control for in between- syndrome comparisons, and which variables may contribute to within- syndrome variability. The important variables to assess may be different for different genetic disorders.

3.5.3 Sample recruitment

Many studies on behavioural phenotypes have recruited their samples via parent support groups. The advantage to this method of recruitment is that large samples can be obtained which is often difficult when research is being conducted on rare syndromes. Also, parents’ associations are often interested in research and tend to be highly motivated to take part (O'Brien & Yule, 1995b) resulting in high response rates. However, there are a number of disadvantages to recruiting the sample in this way. Not all individuals in the parent association will have a confirmed diagnosis of the genetic syndrome, and independent validation of the diagnosis may be necessary. It cannot be assumed that those who are members of the parent association are entirely representative of total population of individuals with that genetic syndrome. Recruiting in this manner may lead to the over-reporting of behaviours. It is possible that parents may have read literature describing

behaviours associated with the syndrome and over-report the presence of these characteristics in their child.

There are several other places where individuals with genetic syndromes could be recruited including clinical services for individuals with particular genetic syndromes and genetic registries. These methods of recruitment also have their drawbacks. Those cases attending clinical services are likely to be more severely affected than those who do not contact specialised services. Genetic registries are useful to recruit individuals with disorders where a genetic test is available. Recruitment from such a registry would then ensure the diagnosis of the syndrome in all cases. Ideally, samples for behavioural phenotype research would be recruited from epidemiological studies. However, this would be a massive undertaking for research on any specific syndrome as most are exceedingly rare. Therefore, it may be necessary to rely on other recruitment methods,

3.5.4 Measurement procedures

Investigation of a behavioural phenotype may involve the assessment of behavioural, psychiatric or developmental disorders, or general profile or specific aspects of cognitive ability in individuals with the genetic syndrome, (see section 3.5.1). For each of these potential lines of enquiry there are many different assessment methods from which to choose, some of these are discussed below.

Well-validated measures are available to assess behavioural disorder, developmental and psychiatric disorder, and cognitive abilities. Using a well- validated instrument has a number of advantages over developing one for the purpose of the study. Measurement error can be minimised. The psychometric properties of the instrument are known and definitions for the behaviours described in the instrument are provided. To follow are a few examples of the measures available to assess general behavioural disorder.

The Child Behaviour Checklist (CBL; Achenbach, 1991) has a number of different versions: two parental questionnaires, one for children aged 2-3

years and one for children aged 4-18 years, a teacher questionnaire for children aged 4-18 years, and a self-report questionnaire for children aged 11-18 years. The questionnaires were designed to discriminate between children adapting successfully and those deemed to have behavioural and emotional problems. The CBL manual provides normative data for a non­ clinic sample (children that had not received mental health services or special remedial school classes in the last 12 months) and a clinic sample (children who have been referred to mental health services for behavioural and emotional problems). The CBL was not designed for use with children with learning disabilities and is therefore more applicable to behavioural phenotype research where there is little associated intellectual impairment.

The Developmental Behaviour Checklist (DBC; Einfeld & Tonge, 1994) was developed specifically for use with children with learning disabilities aged between 4-18 years. It has a parent and teacher version. The items were determined by examining the prevalence of the behaviours characteristic of behavioural and emotional disturbance in children with mild, moderate, severe and profound learning disabilities. The manual accompanying the DBC provides normative data for each of these ability groups.

The Aberrant Behavior Checklist (AbBC; Aman & Singh, 1986) was developed for use with adults with learning disabilities. The original version was developed by determining the prevalence of behaviours characteristic of behavioural and emotional disturbance in adults living in institutions. A further version was developed for adults living in the community. The AbBC community manual (Aman & Singh, 1994) provides data for adults with mild, moderate, severe, and profound learning disabilities. These and other questionnaires measuring general behaviour disorder are frequently used as the first stage in delineating the behaviours associated with particular genetic disorder.

To assess psychiatric and developmental disorder one can apply the diagnostic criteria outlined in the DSM or ICD diagnostic manuals. The

problem with this approach is that many individuals with genetic syndromes also have learning disabilities. The validity of applying the diagnostic criteria to individuals with learning disabilities is not currently known. To apply the diagnostic criteria assumes that these individuals suffer from the same psychiatric disorders as non-learning disabled individuals (Dykens, 2000). This may not be the case, children with learning disabilities may show symptoms of psychiatric disorder but they might manifest in a different way. For example anxiety is likely to manifest differently in someone without speech than in someone who is able to speak.

Another way of investigating the presence of psychiatric and developmental disorders in genetic syndromes is to use a rating scales or interviews focused on individuals with learning disabilities. For example, a version of the Reiss Screen (Reiss, 1990) was designed to assess behaviours characteristic of psychiatric disorder in children with learning disabilities aged between 4 and 21 years. It provides scores on 10 sub-scales developed using samples of individuals with mild, moderate, severe and profound learning disabilities. The 10 scales are reported to be closely related to the DSM-III-R diagnostic criteria for psychiatric disorders. The Autism Diagnostic Interview (ADI; Lord, Rutter, & Le Couteur, 1994), the Childhood Autism Rating Scale (CARS; Schopler, Reich 1er, DeVellis, & Daly, 1980) and the Autism Behavior Checklist (ABC; Krug et al., 1980) all assess the presence of autistic behaviours and are suitable for use with children with learning disabilities. The ADI is a parental interview, the ABC is a parental questionnaire and the CARS is an observation schedule.

To investigate cognitive profiles associated with a genetic syndrome, one can employ the use of standardised IQ tests such as the Wechsler Intelligence scale for Children (Wechsler, 1991) or the Wechsler Adult Intelligence Scale (Wechsler, 1986). Investigation of scoring on the sub-scales of these IQ tests can provide insight into the particular strengths and weakness of cognitive abilities. To investigate more detailed cognitive processes, it may be

necessary to use more specific tests, e.g., tests of face recognition or spatial rotation.

While the use of well-validated measures is advantageous over using newly designed measures; their use may pose a problem for one particular line of enquiry in behavioural phenotype research: general behaviour disorder. Instruments such as the CBL or the DBC have been developed by assessing the prevalence of particular behaviours in large populations of individuals (e.g., for the CBL, children referred to clinics for behavioural problems, for the DBC, children with learning disabilities in general). Genetic syndromes are rare and so any behaviour that is specific to one particular syndrome is likely to have had a low frequency within the test samples and would have been excluded from the final versions of the checklists. Thus, if one of these instruments were used in isolation to determine syndrome-specific behaviours, many such behaviours might be missed. One way to overcome this difficulty would be to develop a syndrome-specific measure to use in conjunction with the well-validated measure as suggested by Rosen (1993). For example, a syndrome-specific questionnaire could include items describing the entire range of behaviours seen in individuals with the syndrome of interest. A cluster analyses could then be used to provide some construct validity for the behavioural phenotypes measured (Dykens, 1995).

A method encouraged in the behavioural phenotype field has been to develop one questionnaire that can be used to compare behaviours across syndromes. It is unlikely that these questionnaires would contain enough items to describe all the behaviours associated with any particular syndrome. As with the well-validated measures of behavioural disorder, sole use of such a questionnaire may lead to some syndrome-specific behaviour going unreported. However, these questionnaires would be very useful to compare syndrome against syndrome. They could be employed to investigate which behaviours may be common to one or more disorders, (partial-specificity effects), and investigate which behaviours are specific to just one genetic syndrome (total-specificity effects). The Society for the Study of Behavioural