7.2 Validity of a construct
7.2.7 The future of validating psychiatric disorders: Towards DSM-5 and beyond
The current American classification schema (the Diagnostic and Statistical Manual of Mental Disor- ders, 4th edition, Text Revision(DSM-IV), American Psychiatric Association [23]) as it was originally con- ceived, has an uncertain future given that it was derived from clinical consensus to address primarily the need for diagnostic reliability rather than valid- ity [24], and has not yet been reoriented towards state-of-the-art investigations of the aetiology of psychiatric disorders. The shift in perspective to inclusion of aetiologic information is, in part, the result of a disappointing lack of identifying unequiv- ocal, consistently replicated susceptibility genes for these disorders (as currently defined in the DSM), despite high heritability estimates and the develop- ment of powerful research tools such as genome-wide association studies [25, 26].
This may not be surprising given that the organisation of the DSM classification is not based on the pathogenesis of disorders, but rather on operationalised sets of categorical criteria based on signs and symptoms from clinical observation and research [24, 27, 28]. As pointed out by Steven Hyman,
Without genotypes, objective tests, clues to pathogenesis and even adequate family and longitudinal studies, it was not possible to establish a true empirical base for valid diagnoses in DSM-III, DSM-II-R, DSM-IV and
DSM-IV-TR [24, p. xiii].
Furthermore, Hyman [24] and other investigators have voiced concerns over how the DSM defines the thresholds for categorising disorders by delineat- ing arbitrary cut-off points for normally distributed variables, such as behaviour traits, and for contin- uous measures such as severity and chronicity of the illness. Hyman [29] notes that many patients do not fit precisely into these categories and hence the DSM has relied extensively on the catch-all term
‘not otherwise specified (NOS)’. In fact, the NOS diagnoses are more commonly used than a number of the specifically named disorders. The DSM, now
under revision, is expected to address the validity of the diagnostic system by including experimental criterion sets aimed at incorporating new genetic and neurobiological findings in its fifth edition [29]. This is a shift in approach from the use of categories based on clinical syndromes and levels of functioning [30]
to biologically valid phenotypes that can potentially address questions concerning illness aetiology and clinical treatment.
In general, psychiatric disorders are likely to be characterised by complex multifactorial and polygenetic aetiologies marked by the interaction of numerous genes with each other and a wide range of environmental risk factors, resulting in varying phenotypic expression from normal to clinical relevance [31]. Although we know there are high heritability estimates for many psychiatric disorders, single genes with sufficiently large effects are not likely to generate most disease phenotypes. Instead, the genetic contribution to psychiatric disorders can be viewed as the combined effect of a number of different genes, each with a small or moderate effect on disease liability [25]. Individually, each gene may have only a slight effect on the phenotype such that close relatives may share several susceptibility variants, although one relative may develop the disorder and another may not [25].
Environmental factors also play a significant role in the aetiology of psychiatric disorders, for example as epigenetic factors (i.e. exogenous exposures that influence the expression of genes). As illustrated by research findings on schizophrenia and MDD, environmental influences have included early foetal or neonatal events, such as exposure to obstetric complications [32–35], viruses [36–39], poor nutri- tion [40, 41] social conditions such as living in urban compared with rural regions [42, 43], and migration [44–47]. One example of environmental factors that has influenced the expression of genetic polymorphisms is the increased risk for schizophre- nia in individuals who both smoke cannabis and have a functional polymorphism in the catechol- O-methyltransferase (COMT) gene, a gene respon- sible for metabolism of dopamine [48]. Another example is the role of the serotonin transporter gene-linked polymorphic region (5-HTTLPR, 5-HTT (5-hydroxytryptamine transporter) gene-linked poly- morphic region), which appears to increase the risk
for MDD only among those carrying the short ‘s’
allele and in the context of stressful life events, such as early childhood trauma [49].
Given the complexity of finding genetic causes of psychiatric disorders per se, there has been a surge of research focused on using intermediate phenotypes or traits in genetic modelling of disorders, called endophenotypes [50]. Endophenotypes are quanti- tative or continuous traits found more commonly in psychiatrically ill individuals and their unaffected family members (i.e. family members not meeting the same psychiatric diagnostic criteria) than in the healthy population. Endophenotypes are hypothe- sised to underlie or precede disease onset or the expression of a clinical phenotype (i.e. as measured on a continuum of the aetiologic pathway to the clinical phenotype), and are assumed to be strongly associated with the expression of genes that underlie the disorder [30, 50, 51]. The rationale behind using endophenotypes in molecular genetics research is that (i) Traits represent more elementary phenomena of decreased complexity than the clinical phenotype and thus will likely have stronger associations with spe- cific functions of genes and hence be more genetically informative (i.e. the phenotype will segregate with the susceptibility locus) [26, 30], and (ii) by includ- ing unaffected family members, endophenotypes may afford the investigator greater power to detect link- age than a categorical diagnostic approach [26].
While the endophenotype approach is now widely used, the identification of quantitative endopheno- typic traits of the clinical phenotypes remains contro- versial [52], given that it is still unclear how informa- tive they are in contrast to the DSM categories [26].
Thus criteria for evaluating the validity and utility of endophenotypic markers for research in psychiatric genetics have been proposed by several investigative teams, including Gottesman and Gould [30], Skuse [53], Doyleet al. [54] and Waldman [55]. Based on the guidelines set forth in the psychiatric literature, Bearden and Freimer [51], have proposed a set of criteria viewed as both necessary and sufficient.
1 Endophenotypes should be familial, with at least moderate heritability, and should be detectable in those with the mental illness associated with the phenotype as well as in unaffected family members.
2 Endophenotypes should be part of the casual chain in the relationship between genes and the DSM diagnosis rather than an effect or sequelae of the disorder.
3 Endophenotypes should be reliable (internal consistency), and have test–retest reliability (at least within a particular clinical state, and prefer- ably across clinical states in illnesses with an episodic pattern), sound psychometric properties (e.g. can discriminate across a broad range of individual differences) and good concurrent validity (convergent and divergent validity) with respect to hypothesised endophenotypes.
4 Endophenotypic traits should exhibit a contin- uous distribution (ideally, normally distributed) within the general population.
5 The endophenotype should be associated with an increased risk for a particular DSM diagnosis.
Illness-specificity is desirable but not required.
The authors also add that an ‘optimally infor- mative candidate endophenotype should: (i) relate to reasonably well-characterised neural systems models, and (ii) involve homologies of expression across species (to enable development of animal models).’ [51, p. 309]
The use of quantitative endophenotypic traits for understanding the genetic nature of psychiatric dis- orders fits well with the fact that there is a high comorbidity of psychiatric illnesses not only among psychiatric illnesses but with general medical disor- ders [56, 57]. In fact, only 10–20% of lifetime diag- noses are single disorders [58]. For example, Tsuang andet al. [59] point out that many genetic studies in the late 1990s failed to show linkage to schizophrenia based on a DSM diagnosis of schizophrenia alone, but found stronger linkage when the phenotype was broadened to include additional psychotic disorders (e.g. [60, 61]). As another example, chromosome 13q [62–64], 4p [65], 22q [62, 63] and 18p [62, 66]
have been implicated as promising genomic regions for schizophreniaandbipolar disorder. Thus, for an understanding of the nature of schizophrenia and other psychotic disorders, these examples illustrate the relevance of traits that are shared across psychotic disorders and must be distinguished from the traits that may be specific for the disorders themselves.
In psychiatry, genetic studies using endopheno- types have been met with moderate success, including the development of animal models based on these traits [30, 67]. Endophenotypes in psychiatry have been described for several disorders including schizophrenia, mood disorders, Alzheimer’s disease and personality disorders. Schizotaxia, for example, is a clinical condition that indicates a predisposition or liability, to schizophrenia. A concept first termed by Meehl in 1962 [68] and which has subsequently been reformulated to reflect current research [69], schizotaxia is a more subtle brain disorder than schizophrenia, marked by negative symptoms and neuropsychologic impairment. A number of non-psychotic, first-degree relatives of persons with schizophrenia exhibit clinical and neurobiological abnormalities that are also manifest in patients with schizophrenia [70]. Family studies indicate that schizotaxia is present in about 20–50%
of non-psychotic adult relatives of persons with schizophrenia [71, 72], with about 10% of relatives developing psychosis and another 10% developing schizotypal personality disorder [73].
Schizotaxia may well express the aetiologic mech- anisms that underpin schizophrenia more clearly than the clinical symptoms of the disorder [70].
For example, Tsuang et al. [69, 74] conducted a validation study of schizotaxia based on the treat- ment of nonpsychotic, adult first-degree relatives of patients with schizophrenia using the antipsychotic medication risperidone, a drug which has been found to ameliorate negative symptoms and neuropsycho- logic abnormalities in persons with schizophrenia.
The authors hypothesised that if schizotaxia is bio- logically related to schizophrenia, the negative symp- toms and neuropsychologic abnormalities in the schizotaxic relatives would improve with risperidone treatment. In this study, all study subjects exhibited moderate levels of negative symptoms and neuropsy- chological deficits at baseline, and after a 6-week, open-label course of risperidone, these symptoms and cognitive deficits improved in five of the six rel- atives supporting the authors’ hypothesis in terms of predictive validity. Though the findings were only preliminary, they suggested common aetiologic ele- ments between the two disorders [69]. The authors later published a study of the concurrent validity of schizotaxia in a group of 27 adult first-degree
relatives of patients with schizophrenia [75]. Of these subjects, eight individuals met criteria for schizo- taxia and were compared with 19 control subjects who were free of DSM-IV psychiatric diagnoses.
The authors found that in contrast to those without schizotaxia, the schizotaxia group exhibited signifi- cantly lower levels of functioning and had a lifetime substance abuse diagnosis rate (50%) similar to that among persons with schizophrenia. Findings in this study provided further validation of schizotaxia as a psychiatrically relevant, familially-related condition closely associated with and aetiologically related to schizophrenia.
In summary, new approaches to validating psy- chiatric diagnoses are being developed to incor- porate aetiologic information with regard to the psychiatric disorder rather than relying on symp- tom and functioning information alone. The focus here on endophenotypes illustrates this trend and reflects the expressed need to identify characteristics that are likely related to specific genetic traits and other biomarkers for psychiatric illnesses. However, the search for traits associated with the underly- ing biomarkers for the illness is still in its infancy.
Given the new genetic methodologies and biomed- ical imaging technologies, there is a realistic hope that future classification systems beyond DSM-5 will include specific biomarkers underlying these illnesses that may help tailor specific treatments to affected individuals in a reliable and valid manner.