The moving target of psychiatric diagnosis

Abstract

Clinical diagnosis and/or phenomenological description is the cornerstone of both clinical care and research in psych\xad iatry. Clinical management and guidelines are structured around individual diagnoses (e.g., major depression, schizo\xad phrenia, bipolar disorder), with the implicit understanding that such diagnoses mean the same to most clinicians (reli\xad ability) and that they have predictive value (validity). Another relevant aspect of diagnosis is its clinical utility — a measure that is difficult to estimate correctly. However, indi\xad rect data, such as increasing burden of psychiatric disorders, might be in part explained by diagnostic limitations.1 Similar considerations apply to research. While molecular, genetic, and brain imaging technologies have been making considerable progress, the phenotype definition frequently becomes a bottleneck with multiple ramifications: inconsis\xad tencies and difficulties in properly defining the phenotype lead to seemingly contradictory findings. It is my hope, how\xad ever, that a careful examination of such discrepancies can be instructive and help move psychiatric research forward. Let us consider some examples. An analysis of symptoms and life trajectories led to a proposal for a general psycho\xad pathology “p factor.”2 This p factor has been viewed as a single dimension underpinning the various psychiatric symptoms and explaining a large portion of interindividual variability with respect to presence of psychopathology. In parallel, re\xad sults of genome\xadwide association studies led to a proposal of a common genomic p factor for psychiatric disorders.3 Similarly, the concept of general neurotic syndrome, while absent from current classification schemes, has existed for some time and is supported, for example, by twin data.4 Contrary to these con\xad cepts, in a family\xadbased study of a Swedish population, Kendler and colleagues were able to separate rather clearly the 3 major psychiatric disorders (depression, bipolar disorder and schizophrenia).5 From family history data in health population registries, they constructed family genetic risk scores and com\xad pared them across the diagnoses. Each of the 3 conditions had minimal overlap except for elevated family genetic risk scores for depression in those with bipolar disorder, which is unsur\xad prising given that finding’s consistency with the results of older family studies. In an extended analysis that included a number of other diagnoses, the separation became less distinct, showing correlations between, for example, major depression and anxiety or substance use and attention\xaddeficit/ hyperactvity disorder (ADHD).6 As of now, some of the largest genetic studies in psych\xad iatry, counting tens or even hundreds of thousands of cases and hundreds of thousands controls, have identified at least 270 susceptibility loci for schizophrenia (preprint evidence: https://www.medrxiv.org/content/10.1101/2020.09.12.2019 2922v1), 64 for bipolar disorder,7 102 for major depression,8 as well as multiple loci for other diagnoses. Many of the genetic variants implicated in one condition seem to be involved in other diagnoses, leading to high genetic correlations between individual psychiatric diagnoses and suggesting shared herit ability.9 Several recent genome\xadwide association studies found a substantial correlation between the genetic liabilities for most main psychiatric diagnoses (schizophrenia, bipolar disorder, major depressive disorder, autism, ADHD, or anx\xad iety disorders). These results were obtained using mostly criteria\xadbased diagnoses without any further refinement. Sev\xad eral investigations went even further in simplifying their re\xad cruitment and relied on limited phenotyping and/or self\xadre\xad ported diagnoses. Studies of major depression or bipolar disorder in self\xadidentified individuals in databases such as 23andMe or the UK Biobank resulted in considerably weaker genetic associations and low so\xadcalled single nucleotide poly\xad morphism (SNP) heritability (proportion of phenotypic vari\xad ance explained by markers identified by genome\xadwide asso\xad ciation analysis).7,10 Conversely, a recent study of the genetics of patients with major depressive disorder treated with elec\xad troconvulsive therapy (ECT) — arguably a much more severe and more specific diagnosis — yielded SNP heritability about 4 times greater than a standard genome\xadwide association study of major depression.11 In addition to the core psychiatric diagnosis, other fea\xad tures come into play. For instance, while structural brain imaging studies of bipolar disorder found reproducible dif\xad ferences from healthy unaffected individuals, it appears that some of the findings are not due to the disorder itself, but rather are secondary to metabolic factors such as obesity or insulin resistance that are frequently comorbid with severe mental illness.12