Blood Brothers or Just Neighbours: Modelling the Functional Psychoses

The proposition that schizophrenia and bipolar disorder are a single condition

A recent Lancet paper attracted considerable media attention, being interpreted by many journalists as demonstrating that schizophrenia and bipolar disorder were effectively a single condition [1]. In essence, the paper challenged the long-held binary or dichotomous model of the functional psychoses.

The Lancet article can be briefly overviewed. Analysing case register data involving more than 2 million Swedish individuals, Lichtenstein et al. identified those who had been hospitalized (on at least two occasions) and discharged with a diagnosis of schizophrenia or bipolar disorder, and then quantified the risk of each of those conditions in relatives of those probands with age- and gender-matched unaffected individuals [1]. They demonstrated that probands with schizophrenia were likely to have relatives with higher rates of schizophrenia and of bipolar disorder and, similarly, that those probands with bipolar disorder were likely to have relatives with higher rates of bipolar disorder and of schizophrenia. The authors’ abstract noted that ‘Shared environmental effects were small but substantial…for both disorders’ and that ‘comorbidity between disorders was mainly (63%) due to additive genetic effects common to both disorders’.

In the Discussion section the authors extrapolated beyond the data set, arguing that their results are ‘in line with previous….studies that suggest a common genetic contribution for schizophrenia and bipolar disorder’. Together with a second statement: ‘schizophrenia and bipolar disorder share common genetic causes’ and the authors’ claim in the final paragraph that their results challenge ‘the nosological dichotomy between schizophrenia and bipolar disorder’, it was not surprising that the media interpreted these findings as demonstrating that schizophrenia and bipolar disorder were essentially the same entity.

Such authoritative statements in the paper – and in an accompanying editorial [2] – move beyond the tenor of the paper's Introduction, where a more cautious approach is evident. For example, the authors suggest there that ‘some genes are probably associated with the risk for both disorders and some with the risk for only one disorder’. Such caution is effectively expunged in their Abstract and concluding paragraphs – and in the accompanying editorial subtitled ‘time to face the future’ [2]. The editorial noted Kraepelin's splitting of the two ‘functional’ psychoses more than a century ago [3] and that it has been subsequently ‘but erroneously’ assumed that the ‘two disorders are discrete, natural disease entities with distinct pathogenesis’, before the editorial authors proposed that ‘We must now ask whether clinical practice and research can continue to be best served by persistence in basing our diagnoses on the binary concept’. The editorial went on to contemplate whether ‘the functional psychoses are better conceptualized as a continuum’ – as against a ‘set of overlapping pathological processes’. Such a unitary model (effectively consolidating the two psychotic states) is not just an isolated speculation – such a proposal is currently being considered by a DSM-V committee.

Now, let me put the contrary argument, against the data from the Lancet paper and its purported evidence to support the merging of these two diagnoses. Let us consider the clinical neighbourhood. Most clinicians view schizophrenia and bipolar disorder as (commonly but not invariably) distinguishable. The distinctions emerge from differing cross-sectional clinical features and natural histories; and from differential response to certain medications. The comparatively superior outcome for those with a bipolar disorder (in terms of periods of normative functioning, greater likelihood of being able to maintain family relationships and to hold down a job) is even recognized by many patients, who bluntly acknowledge that they would much prefer a diagnosis of bipolar disorder. True, some patients with seemingly true bipolar disorder will have a poor outcome or show cognitive decline on psychometric testing. Presumably, their progress reflects manifestation of a genetic cascade that parachutes them into the no man's land between the purer expressions of each disorder. True, also, that there has been a wide body of high-quality work (including genetic and brain imaging studies as well as phenomenological analyses) that has failed to find definitive support for the binary view. Further, we are increasingly recognizing that some patients with a bipolar disorder will obtain benefit from antipsychotic drugs, whether as monotherapy or as an augmentation strategy. The last, however, is only a limited argument for the unitary model, because antipsychotic drugs can be very broad-brush in their effects, even helping patients with anxiety, anorexia nervosa and borderline personality disorder. This evidence is further limited by the reality that some physical treatments that are helpful for bipolar disorder (e.g. lithium; electroconvulsive therapy for bipolar depression and even severe psychotic mania) are rarely beneficial for schizophrenia.

Thus, a clinician operating to a binary model should be able to accurately diagnose schizophrenia and bipolar disorder in a high percentage of assessments, but also recognize that some patients will show shared features that disallow ready binary assignment – a situation possibly handled by using an interregnum schizoaffective diagnosis, and then regularly reviewing the diagnosis over time. For those with an identifiable higher order diagnosis, subtyping might then be relevant (e.g. determining whether an individual had a bipolar I or bipolar II disorder), as again there are likely to be meaningful management nuances to such subtyping.

Clinical anomie seeks certainties

I suspect homogenizing (of psychoses and of depressive conditions) reflects, in part, a sense of nihilism held by many in the profession. Current clinical nihilism is well captured in an introspective piece published in January 2009 in the American Journal of Psychiatry entitled ‘Why psychiatry is the hardest specialty’ [4]. Here the psychiatrist author details how in clinical practice she does not ‘know what's wrong’ nor ‘what I'm doing’, confused by both the perceived ineffability of clinical practice and even further by the placebo-controlled trials that prevent her from knowing whether she is really helping or just imagining that she is helping.

For those who seek certainty, an inability to tolerate ambiguities is often resolved by adopting a simple model. In relationship to psychiatric classification, any perceived failure to conclusively prove that schizophrenia and bipolar disorder, or that melancholic and non-melancholic depression cannot be pristinely and invariably distinguished, is evidence of a general tendency to resolve the dilemma by dimensionalizing some component such as severity, persistence or chronicity. The alternative approach, that is, seeking to define the commonalities and the differences, is more fruitful, especially when the issue is substantive, and new definitional models are then developed. Anthropologists are comfortable with ‘thick description’, and medicine has constantly progressed by facing such challenges. Kendell noted how ‘the pox’ and ‘dropsy’ were once viewed as unitary conditions, and how progress occurred when the pox was divided into chicken pox and smallpox (allowing quite differing prognoses), and when renal and cardiac causes of the dropsy were identified, allowing quite differing management approaches for those differing aetiological expressions [5]. Medicine proceeds by progressively subtyping (e.g. diabetes, hepatitis), with progressively identified causes shaping differential management strategies largely based on pattern analyses of clinical subtypes. The trend in psychiatry, however, is to dimensionalize: currently the depressive disorders; in future the personality disorders; and, seemingly, now the functional psychoses.

Why the Lancet article is less definitive than it seems

Let us return to the Lancet article. First, the authors acknowledge that they analysed ‘non-standardized diagnoses’, so allowing diagnostic biases. The well-cited US–UK diagnostic study (in which videotapes of hospitalized patients were nine times more likely to generate a diagnosis of schizophrenia from US as compared to UK psychiatrists) illustrates the risk of diagnostic bias [6]. Every misdiagnosed proband (i.e. false-positive schizophrenia, false-positive bipolar disorder) in the Lancet article would have shaped the direction of their findings to support a unitary model.

Another methodological point. The Swedish patients had all been hospitalized. This effectively excluded those with milder bipolar II conditions (i.e. the bipolar majority – and in whom schizophrenic features are extremely rare), and weighted the sample to those who, as a consequence of being psychotic and hospitalized, were more likely to have severer, regressive and even schizoaffective expressions.

Turning to analytic interpretation, although a number of proband–relative analyses (e.g. involving offspring, siblings, maternal and paternal half-siblings as relatives) were undertaken, trends were similar and can best be illustrated by summarizing the parent–offspring data. Here, if the proband had received a diagnosis of schizophrenia, the risks of schizophrenia and bipolar disorder in a relative were 9.9 and 5.2, respectively. Conversely, if the proband had received a diagnosis of bipolar disorder, the risks of bipolar disorder and of schizophrenia in a relative were 6.4 and 2.4. Whether the biological relationship was parent–offspring, sibling–sibling, sibling–maternal half-sibling or sibling–paternal half-sibling, relative risks quantified the more likely psychotic condition in a relative as the one diagnosed in the proband. In essence, the consistent trend actually allows a ‘breeding true’ interpretation – but this is neither recognized nor contemplated by the authors.

The most substantive concerns, however, are evident in the authors’ interpretation and emphases. They initially concede (in their Introduction) the possibility that a ‘shared causative risk factor is a risk factor for part of each disorder’. In essence, this model allows that certain (more diagnostically non-specific) genes may dispose to psychosis and that other (disorder-specific) genes may contribute to differing phenotypic expressions (that would shape the differing phenotypic expressions of schizophrenia and bipolar disorder). For the authors to have argued that their paper finds support for the first proposition – and to so limit their conclusion – would have been valid. Their implication – that ‘shared genes’ provide the whole story – was the conclusion picked up by the media, and risks reification by those who wish to diagnostically assign schizophrenia and bipolar disorder to the same category.

More likely explanations of the cognate expression of both disorders (if a genetic cause is to be accepted) include that schizophrenia and bipolar disorder are influenced by both shared and non-shared genes, or that a single gene or set of genes has pleotropic expression across the two disorders. An example of the first phenomenon is of the genetic susceptibility to both type I diabetes and to celiac disease [7] contributed to by common alleles. The second phenomenon involves ‘a single gene contributing to multiple phenotypic traits’ [8]. As noted by Schulze and McMahon, ‘there are many examples of diseases with distinct symptoms, course, and treatment that share genetic factors’ [9]. As an example, they detailed how the B27 haplotype of the human leucocyte antigen is a strong risk factor for (i) ankylosing spondylitis in some individuals; (ii) anterior uveitis in others; and (iii) Reiters syndrome in others, but that ‘each of these autoimmune diseases affects a different organ system, produces specific signs and symptoms, and responds best to specific treatments’.

Finally, if it were to be shown (say) that sea urchins and humans share genes, would we assume shared categorical status? It is of relevance then to note that a paper in Science[10] did identify that many gene families are shared across sea urchins and humans – but here the authors did not conclude that equivalence of genotype extrapolated to equivalence of phenotype.

Some conclusions and implications

The bottom line – and an interpretation ignored by the Lancet authors – but one that is central to any consideration of this issue–is that, even if there was a single pristine gene causing both ‘schizophrenia’ and ‘bipolar disorder’, this would not alone argue for – or establish – a unitary model. Further, logic would dictate that management should be directed at the phenotypic condition (e.g. uveitis vs scoliosis, schizophrenia vs bipolar disorder) rather than at any genotypic contribution – unless perhaps a preventative or prophylactic model is the issue.

If a unitary model is accepted, then its simplicity invites all the research and clinical limitations associated with a non-specific construct (such as ‘major depression’), and where the true heterogeneity of the constituent disorders (here schizophrenia, schizoaffective disorder, bipolar I, bipolar II) is ignored and management models then assume some form of treatment as having universal relevance. One specific risk, for example, is of antipsychotic drugs being positioned as a universal treatment, whether by theoreticians and/or by the pharmaceutical industry. This is not to deny that such drugs may be of benefit to many individuals with those constituent disorders but it is doubtful that they have universal application, while their side-effects profiles are often substantive. By contrast, a subtyping model is more challenging but, by embracing and pursuing the complexities, aligns psychiatry closer to the science of medicine.

The Lancet article is best interpreted as demonstrating that some genes may contribute to certain expressions of both schizophrenia and bipolar disorder. It does not demonstrate a unitary model for the psychoses, as many interpret that article. Research and clinical management are unlikely to be advanced by embracing a simplistic dimensional or continuum model for schizophrenia and bipolar disorder.

On reflection, it seems a no brainer: divided we stand, united we fall.