Email Icon Facebook icon Twitter Icon GooglePlus Icon Contact

User Top Menu

Searching for Useful Schizophrenia Subtypes

10 Jan 2017

by Michele Solis

Schizophrenia takes on different forms in different people, and many psychiatry researchers feel that this heterogeneity is a major obstacle to better understanding and treatment of the disorder. Scientists have hypothesized that there may be different underlying biological reasons for any one person’s illness.

To try to glean some organizing principles for subgroups of patients, early psychiatrists relied on observable symptoms that the patient reported or signs that the clinician detected, and through the years came up with subtypes such as paranoid, disorganized, or residual schizophrenia. But these have largely fallen by the wayside, and were struck from the latest version of the Diagnostic and Statistical Manual of Mental Disorders (DSM). The problem is that they weren’t very useful.

“The only reason to have subtypes is that you would need them for treatment or for predictive or for other clinical reasons,” said Carol Tamminga of the University of Texas Southwestern in Dallas.

Throughout the 20th century, researchers also looked to biology for direction, but didn't find anything informative in blood samples or postmortem brains. As methods to examine the human brain and genome have multiplied over the past decade, however, researchers are seizing upon different measures in attempts to make sense of schizophrenia’s heterogeneity. Finding biologically based subtypes may point to different etiologies, predict disease course, help get the right treatments to the right patients, as well as spur the development of new drugs.

“There's probably several different biologies underlying the symptoms we see in our patients,” said Oliver Howes of King’s College London, UK. “I think in the past there's been a one-size-fits-all approach, and I think we've begun to realize that isn't the case,” he said.

Meaningful subtypes so far remain elusive, though there is no shortage of candidates. A recent study has argued for subdividing schizophrenia according to the level of brain muscarinic acetylcholine receptors and associated gene expression patterns (Scarr et al., 2016); others have found that a subset shows signs of brain inflammation (Fillman et al., 2014), while still others have tried to partition schizophrenia according to genetic variance, with controversial results (see SRF story here). Combining multiple measures of biology may reliably identify a subtype, but getting to a meaningful one will still require linking it to a particular treatment or outcome, experts say.

“Sometimes people use subtype just to mean that there's a biological characteristic that differentiates a single group. That's all well and good, but it's kind of a meaningless subtype unless you can tie a therapy or a pathophysiology to it,” Tamminga said.

In this article, we look at three recent examples, including work that has identified subtypes based on cognitive changes and brain scans; evidence that treatment-resistant schizophrenia (TRS) has a qualitatively different biology from cases that respond well to drugs; and an effort that has combined multiple brain measures to obtain “biotypes” that transcend the schizophrenia diagnosis.

Cognitive combinations

Though not a direct measure of biology per se, changes in cognition may reflect shifts in brain physiology. A recent study in JAMA Psychiatry published November 9, 2016, supports this idea, as it found characteristic brain differences in people with schizophrenia that corresponded to their cognition changes with illness (Weinberg et al., 2016).

The work confirmed IQ-based groupings first identified over a decade ago: people who had preserved cognition compared to their abilities before they were ill, those whose cognition declined with illness, and those who had impaired cognition to begin with, which then did not worsen (Weickert et al., 2000).

The new study also obtained brain scans with magnetic resonance imaging (MRI), which revealed reduced volumes in certain brain regions in all people with schizophrenia compared to healthy controls. But the specific patterns of reductions differed among the groups. Among the deteriorated group, the researchers found more widespread volume decreases in those with the most severe cognitive declines, including in the hippocampus, lingual gyrus, and superior temporal sulcus, compared to those with moderate decline. This suggests that the deteriorated group itself may be divided into two subgroups. The brain volumes did not correlate with medication history.

“The results give us a good idea of how we might categorize these patients from a biological perspective, and they might also provide clues as to how some patients may respond to different therapies,” said study leader Thomas Weickert of Neuroscience Research Australia in Sydney.

Though there were substantial differences in the average brain volumes between some subgroups, an individual's brain scan would not necessarily be consistent with the average brain volume to enable assignment to a specific subgroup, Weickert told SRF.

Combining different measures may more cleanly delineate subgroups.

“Linking cognition and brain scan data with other biological aspects of the illness, whether they be genetics or other factors, may provide a more accurate and useful picture of these different subgroups,” he said.

Treatment-resistant type

One potential biologically based subtype for schizophrenia may be a group already well known to clinicians. About one-third of people with schizophrenia are “treatment resistant,” meaning that antipsychotic drugs do not relieve their delusions and hallucinations. These drugs block dopamine 2 receptors (D2Rs) to quell the overactive dopamine signaling thought to drive psychosis. The lack of response in patients with TRS suggests something else is afoot.

“Treatment-resistant schizophrenia is probably the biggest clinical burden because the patients that don't respond to any of our treatments are the sickest, and they spend the longest in hospital, and of course we have the least we can offer them at the moment,” Howes said. “If we understand what's different about the biology of their illness, then hopefully we can develop treatments that actually target the problem and help them.”

The lack of response to antipsychotics found in TRS may make perfect sense. Using positron-emission tomography (PET) to monitor dopamine signaling in the brain, Howes and colleagues found in 2012 that the TRS group did not, in fact, overproduce dopamine; in that case, D2R would be unlikely to reduce psychosis (see SRF story). Instead, elevated glutamate in the anterior cingulate cortex may characterize the brains of people with TRS, as his team reported earlier this year (Mouchlianitis et al., 2016). This might explain why TRS patients can do well on clozapine, an antipsychotic that affects several neurotransmitter systems, including glutamate and serotonin.

Howes has proposed that treatment-resistant patients would fall into a “normo” dopaminergic subtype, which could encompass a glutamate-abnormal type as well as other brain states that have not yet been identified (Howes and Kapur, 2014).

A real test of validity will be whether a brain scan can predict how well an individual will do on an antipsychotic. Howes and his team have completed but not yet published a study that scans people with first-episode psychosis prior to their receiving medication.

“It certainly looks promising from the point of view of dopamine predicting response to dopamine blockers,” he said.

Genetics work may also shore up biological differences between antipsychotic-responsive and TRS groups. Preliminary reports have suggested TRS-specific signals among common variants (see SRF story).

Biological differences such as those found in TRS might explain some drug trial failures. For example, Lilly’s failed mGluR drug (LY2140023) might be more effective in subgroups of patients with confirmed glutamate abnormalities (see SRF story).

Industry is starting to take notice, and to think about including a TRS subtype in drug trials, Howes said. In reality, this gets complicated, however, as subdividing people into smaller groups reduces statistical power. For example, Weickert points out that in an antibody trial he is conducting to treat people with schizophrenia who have signs of ongoing inflammation, many willing patients cannot participate because they do not have the inflammation biomarkers.

“In that case, you have to turn away a substantial number of people who do not meet the strict criteria, which can make recruitment challenging,” he said.

Finally, firming up what exactly constitutes TRS in the first place would also help. A new consensus paper from a consortium offers up a definition after finding that no group was calling TRS the same thing (Howes et al., 2016).

Building biotypes

Biological measures might also blur diagnostic categories if they index something shared by two different disorders. For example, the dopamine overload thought to drive psychosis may be shared in schizophrenia and bipolar disorder with psychosis.

Along these lines, the National Institute of Mental Health’s Research Domain Criteria (RDoC) program aims to reach biologically based descriptions of mental illnesses, and acknowledges that these measures may transect traditional diagnoses.

Consistent with this, the Bipolar Schizophrenia Network on Intermediate Phenotypes (B-SNIP) collaboration headed by Tamminga has identified three subgroupings, or “biotypes,” in a sample of 1,000 people with psychotic illness, based on a battery of phenotyping measures that reflect brain activity including cognition, perception, and eye movement control measures (see SRF story here). Biotype 1 had the most profound cognitive and social impairments, Biotype 2 was marked by heightened sensorimotor reactivity, and the third group was largely normal.

The effort was first conceived as a way to help clinicians tell the differences among a variety of psychotic illnesses, but this did not pan out. When the team discarded diagnosis as a factor, and reclustered the patients based on the biomarker battery only, the three groupings came to light. Each biotype contained people with different diagnoses.

The same approach may well reveal subgroups within schizophrenia, but it will take more data, Tamminga said. B-SNIP is now expanding to phenotype 3,000 more people, and will add genetic data to the mix to see if they can identify a “genetic fingerprint” for the subgroups. The new data may identify new or different clusters across psychotic illnesses or possibly within schizophrenia.

“I think that most of us think that [the current three biotypes] will not be the final word. As we extend these data to include the 3,000 more we are studying now, this will help us refine the battery and the clusters more,” Tamminga said.

But even a very reliable biotype based on multiple measures will need to be tied to treatment or outcome. This will help get the right treatment to the right patients, and may well spur the development of new drugs.

“One of the reasons to identify more biologically homogeneous groups would be to study brain, behavior, and tissues from these groups and find new molecular targets for drugs,” Tamminga said.