The hypothesis of biologically based subtypes of schizophrenia: a 10‐year update

Abstract

A decade ago we proposed that there are two biological subtypes of schizophrenia: type A, characterized by mesostriatal hyperdopaminergia, and type B, without hyperdopaminergia1. One clinical implication was that type A would be associated with good treatment response to antipsychotics, which all act on the dopamine system by blocking dopamine D2 receptors, whereas type B would be associated with poor antipsychotic treatment response1. We also proposed that glutamatergic dysfunction would underlie type B, and suggested additional biologically-based subtypes (C, D, etc.), linked to poor treatment response, yet to be discovered. The hypothesis predicted that biological differences would be present from illness onset. It would be falsified if, for example, there was no response to antipsychotic treatment in schizophrenia despite mesostriatal hyperdopaminergia. Ten years on, we provide here an update on evidence addressing the hypothesis and evaluate the impact it has had. Much of the evidence for dopaminergic differences between treatment-responsive and treatment-resistant groups available ten years ago was potentially confounded by group differences in severity of ongoing symptoms. This has been addressed since then. Positron emission tomography (PET) imaging was used to measure dopamine synthesis capacity (DSC), through radiolabelled 18F-DOPA, comparing people with treatment-responsive schizophrenia taking first-line antipsychotic treatment to people with treatment-resistant schizophrenia whose illness had responded to clozapine (a second-line antipsychotic licensed for treatment resistance)2. When matched for current symptom severity, lower DSC was still observed in the treatment-resistant group, consistent with the hypothesis2. Given that most individuals with treatment-resistant schizophrenia are treatment-resistant from illness onset, a key test of the hypothesis is whether dopaminergic differences are already present from the first episode3. Consistent with the hypothesis, an 18F-DOPA PET study in first-episode psychosis found higher striatal DSC at baseline in patients who went on to show good antipsychotic treatment response, relative to those with poor response. By including people not taking antipsychotic treatment, possible confounding effects of antipsychotic medication on baseline dopamine function were also addressed4. A related methodological approach involves magnetic resonance imaging (MRI) to measure neuromelanin (a by-product of dopamine metabolism) in the midbrain. Using this paradigm in first-episode patients, significantly lower neuromelanin signal was found within the substantia nigra of those who went on to show poor response to treatment relative to those who responded5. Studies have also supported the original proposal that non-response is associated with glutamatergic dysfunction. For example, a study using 1H-magnetic resonance spectroscopy (1H-MRS) found increased glutamate levels in the anterior cingulate cortex in people with treatment-resistant, relative to treatment-responsive, schizophrenia6. Genetic studies in the past ten years have also provided evidence that treatment resistance is associated with enrichment in risk variants affecting glutamatergic signalling pathways, further supporting this aspect of the hypothesis3. It is important to note that not all studies comparing dopaminergic or glutamatergic measures between patients who have responded and those who have not responded to antipsychotic drugs have found differences7. However, as adherence was not established using objective measures (such as antipsychotic plasma levels), it remains unclear if the patients had received adequate treatment. As adequate treatment is a prerequisite to test the hypothesis, poor adherence is a potential confound in a number of studies. One reading of the hypothesis is that the proposed types represent discrete categories. In this case, relevant biological measures should show a bimodal distribution in unselected populations of individuals with schizophrenia. However, evidence to date in unselected populations is more consistent with a unimodal normal distribution, albeit analyses may lack power to detect differences. Notwithstanding the need to test the distribution of data in larger unselected samples, this suggests that the proposed types might potentially be better understood as poles of a continuum. Epidemiological evidence has shown that, in a subset of schizophrenia patients, illness initially shows a good response to treatment, but resistance develops over time3. This is not consistent with a simple A/B dichotomy. The original hypothesis posited that there may be more than one biological subtype associated with treatment resistance, but this finding indicates that some subtypes may develop during the course of illness. This could either be due to iatrogenic effects of treatment, such as D2 receptor upregulation, or changes in the dopaminergic system during the illness that cause breakthrough symptoms3. Other biologically-based categorizations of schizophrenia have been developed, for example using structural imaging measures, and there are examples of subtypes that are unrelated to antipsychotic response8. It is not surprising that a classification based on a biomarker unrelated to the dopaminergic system does not predict treatment response, but these findings highlight that there are aspects of the neurobiology of schizophrenia not covered by our classification. A decade ago, we proposed that biological tests for A and B schizophrenia subtypes could guide treatment decisions and lead to earlier use of clozapine. Since then, 18F-DOPA PET assessment of patients with schizophrenia has shown good predictive power to differentiate responders from non-responders9. Furthermore, it was estimated that, after accounting for the cost of PET imaging, fast-tracking treatment-resistant patients to clozapine in this way would provide a significant reduction in health care expenditure9. More recently, neuromelanin-sensitive MRI was used to identify non-responders and was able to separate them from responders with areas under the curve (AUC) of 0.62-0.855. Alternatively, measuring glutamate in the anterior cingulate cortex through 1H-MRS was able to distinguish antipsychotic non-responders from responders with an AUC of 0.597. These studies show that tests based on biological subtyping are being developed, albeit their translational benefits have yet to be evaluated in routine clinical practice. We also proposed that research into type B schizophrenia could provide insights into its neuropathophysiological underpinnings. In addition to neuroimaging and genetic findings related to glutamate signalling, other biological systems have been investigated. For example, measures of oxidative stress, such as lipid peroxidation, were found to be higher in people with treatment-resistant, as compared to treatment-responsive, schizophrenia3. Thus, a number of biological differences between the subtypes are emerging, which could lead to new therapeutic alternatives for people with type B schizophrenia, up to 60% of whom show limited response even to clozapine3. Whilst evidence accrued in support of the hypothesis during the past decade, some uncertainty remains. This is partly because not all the studies have been positive, although none have disproven the hypothesis and confounds may explain discrepancies, as discussed above. The hypothesis may also need revision to explain the biological basis of treatment resistance that develops during the illness. Notwithstanding these points, there is encouraging evidence of potential clinical utility and cost-effectiveness of tests based on the subtyping. An important avenue for work over the next decade will be to determine their value in clinical practice. Another valuable focus for the future is the further characterization of the neurobiology of type B schizophrenia, which could improve diagnostic approaches further. Critically, this could also identify novel therapeutic targets, and determine if there are further biological subtypes linked to therapeutic outcomes. There is dissatisfaction with psychiatric classificatory systems, such as the DSM, based purely on clinical features. The approach suggested here offers an alternative that links biology to clinical outcomes. We look forward to further testing of the hypothesis and evaluating its translation to clinical practice.

Topics & Concepts