Schizophrenia Research Forum - A Catalyst for Creative Thinking

SIRS 2008—Can Drug Response Be Enhanced in Schizophrenia?

30 June 2008. It was hot in Venice, and it was not just the Mediterranean sun. The first ever Schizophrenia International Research Society conference, held 21-26 June 2008 on the Venetian island of Lido, featured sparks and conflagrations—mostly about the conundrum of schizophrenia genetics. More on that in one of our later reports. Your roving reporter, Hakon Heimer, attended and will report on as many sessions as the espresso could fuel, and we have promises of missives from some of the participating scientists as well. Herewith is our first report.

In a session that took place even before the opening ceremonies, on the afternoon of 21 June, Chairs Anil Malhotra, of Zucker Hillside Hospital, Glen Oaks, New York, and Carol Tamminga, University of Texas Southwestern, in Dallas, helped kick off the conference with a session entitled, "Dissecting the Heterogeneity of Antipsychotic Drug Response: New Approaches to an Old Problem." Although the speakers were hopeful about the possibility that neuroimaging, pharmacogenetics, and other new methodologies would improve the ability of psychiatrists to tailor treatments for different groups of patients, they also seemed to agree that this remains a promise, rather than a reality.

In his opening talk, "The heterogeneity of antipsychotic drug response and prediction of outcome: Are we further ahead?," George Awad of the University of Toronto, Canada, presented a rather sobering historical picture. Sixty years after the advent of antipsychotic drugs (APDs), the situation may be more promising, but the reality is that there are few agreed upon predictors of response to medications. What we have are not biomarkers but some crude clinical predictors, in particular, two, according to Awad—early symptom change and dysphoria. Early symptom change predicts better response, according to work first published by his group several decades ago, and recently supported by a paper from Stefan Leucht and colleagues (Leucht et al., 2008). Conversely, dysphoria reported by patients in the first two hours of treatment portends a cascade including motor and cognitive symptoms.

There have certainly been attempts to find better clinical predictors in schizophrenia, but Awad noted that evaluating these is difficult because investigators have employed different methods in studying sometimes non-comparable study populations. In particular, he singled out shortcomings in the available rating scales—questionable metrics; lengthy, cumbersome, and not particularly sensitive study instruments; too many, poorly validated instruments—and suggested that new assessment scales were needed. Despite this rather sparse résumé, he is optimistic that better predictive ability is in the offing, in part from pharmacogenetics and imaging, but also from better definitions of response and the time framework to achieve this. The "item bank" approach of having a repository of validated assessment has been successful in assessing drug response in depression and anxiety, he said, and should be considered in schizophrenia.

Richard Keefe, of Duke University, Durham, North Carolina, discussed what currently seems like a cul-de-sac in trying to predict the effects of drugs on patients. His talk, "Neurocognitive task performance as an outcome measure in antipsychotic treatment trials," focusing on cognitive assessments of subjects in the Clinical Antipsychotic Trials of Intervention Effectiveness (CATIE), found little to separate any of the typical and atypical drugs from each other. Indeed, the minor improvement seen for these drugs over placebo may signal practice (see SRF related news story) or placebo effects. Although analysis of the CATIE cognition results may be limited by the fact that the study enrolled chronic patients, the European First-Episode Schizophrenia Trial (EUFEST; see SRF related news story) has similar findings, according to unpublished data relayed by Keefe from the authors of that study.

The next major step forward in the attempt to distinguish antipsychotic drugs by their effects on cognition is the Measurement and Treatment Research to Improve Cognition in Schizophrenia (MATRICS) battery of tests. Although there are 10 tests across the seven MATRICS domains (Nuechterlein et al., 2008), Keefe cautioned against assuming that the best dissection of antipsychotic drug effects on cognition would come from selective application of these tests. Composite scores may be less specific but are statistically more powerful, he said. Cognitive analysis of CATIE results showed that there are (modest) correlations between cognitive domains, and that a single composite score describes data best according to modeling (Keefe et al., 2006). He also added that the MATRICS battery should be an improvement over the methods used in CATIE in the way that it takes into account missing data.

Has imaging fared any better as a way to distinguish schizophrenia patients from each other with respect to predicting treatment response? Malhotra's colleague at Zucker Hillside Hospital, Philip Szeszko, described a complex historical picture where structural imaging has served up some candidate structural markers of disease outcome, though they remain unconfirmed. He then described a new MRI study that searched for cortical surface differences between responders and non-responders to risperidone or olanzapine in first-episode schizophrenia, using newer, pattern-matching technologies that can control better for inter-individual differences in the size and shapes of anatomical areas and better assess cortical thickness. Although the researchers report no significant differences in overall gray or white matter, or total volume, they do find that non-responders had greater thinning of occipital and ventral prefrontal cortex than responders.

Moving from structural imaging to functional imaging, Carol Tamminga reported on her group's investigations into medial temporal lobe (MTL) deficits in schizophrenia. Among the questions that she and her colleagues are addressing is whether the antipsychotic effects of dopamine D2 receptor blockers derive in part from "normalization" of neuronal activity in MTL structures. Previous work from her lab has indicated that people with schizophrenia (off medication) show constitutively increased regional cerebral blood flow (rCBF) in the anterior hippocampus, but decreased neuronal activation to memory tasks in this region (as measured by fMRI BOLD). Antipsychotic drugs partially rescue these phenomena, and it may be possible to use measurements of rCBF and neuronal activation to help predict patients' likely responses to medications.

In their current studies, using a task that assesses novelty detection, an early component of encoding new memories, Tamminga and colleagues have found evidence that the same set of abnormalities found in schizophrenia patients in anterior hippocampus is present, but with an even stronger discrepancy between schizophrenia patients and controls, in perirhinal and parahippocampal cortex. Similarly, on memory tasks that measure the flexible use of learned associations, schizophrenia patients off medication show decreased fMRI activation compared with those on medication or normal controls. Tamminga believes aspects of these alterations (those in the CA3 subfield of hippocampus) are related to psychosis, and alterations in dentate gyrus are related to memory per se. As to how these differences are mediated, and how APDs might intervene, Tamminga offered several possibilities—a direct effect of DRD2 and/or D1 receptors in the hippocampus, or indirect effects of dopamine on cholinergic neurotransmission in the hippocampus.

To close out the session, Anil Malhotra addressed the question of whether there has been any movement toward the much vaunted "personalized medicine" in terms of pharmacogenetics? The first generation of studies in this realm focused on whether variation in dopamine or serotonin receptor genes helped determine the likelihood of clinical response to clozapine, but the results were mixed, Malhotra said. More recently, some researchers have moved their work into first-episode patients, in an attempt to avoid the confound of changes that previous APD treatment might introduce for important clinical characteristics such as weight gain. Malhotra's group has reported an association between polymorphisms in the promoter region of DRD2 in a small sample, but larger studies will be needed to confirm and extend this finding. In terms of adverse events, there is evidence for an effect of serotonergic receptor genotype and APD-induced weight gain, and in new work, Malhotra and colleagues have found preliminary support for a role of DRD2 variation in first-episode patients. Their newest work in this regard is a genomewide analysis of treatment response, which has identified a handful of genes with significance below 5 x 106 and which will need additional association evidence for confirmation.—Hakon Heimer.

Comments on Related News


Related News: Psychiatry Aims to Get Personal in Studies of Genes, Drug Efficacy

Comment by:  Anil Malhotra, SRF Advisor
Submitted 1 July 2009
Posted 1 July 2009

Pharmacogenetic studies offer the prospect of the identification of immutable biological predictors of drug response, and may also provide insight into the key molecular mediators of drug efficacy. These studies by Perlis and colleagues and Licinio and colleagues highlight the growth of this field, as well as potential opportunities and challenges associated with this line of inquiry.

Perlis and colleagues have utilized the GWAS approach to dissect the heterogeneity of drug response in the large STEP-BD cohort to identify a number of potential loci associated with response to lithium, and then examined these loci in a separate lithium-treated cohort derived from the U.K. The use of GWAS provides a hypothesis-free comprehensive examination of the genome, and the inclusion of a second sample adds a degree of confidence in the nominally significant GWAS results. The key strength of the study is the large sample size (n = 1,177) and the prospective design of the STEP-BD sample. As the authors note, however, the second sample does not truly represent “replication” as the phenotypes assessed differ between cohorts, with the second cohort restricted to a retrospective global assessment of response. However, it should be realized that the often requested “replication” sample may simply not be feasible with pharmacogenetic designs given the complexities of clinical trial design, and the inherent difficulties in designing and conducting the exact same clinical trial in two similar populations.

This study highlights some of the challenges inherent in pharmacogenetics. In order to adequately power the GWAS analysis, a large sample such as STEP-BD was required, as well as a “replication” cohort because the initial GWAS results would not be likely to meet genomewide significance. The use of this sample does introduce considerable heterogeneity into the dataset (multiple sites, different rating teams, several classes and doses of drugs, differential levels of illness chronicity, variance in prior drug exposure, etc.). Perhaps most notable is that only 23 percent of the subjects actually received lithium as a sole mood stabilizer in the trial, and therefore the signal being detected may reflect additional treatment factors not directly related to lithium. However, this limitation is mitigated by the fact that this mirrors the state of clinical care of bipolar disorder, and the result may therefore be more applicable to the clinical aim of generalized treatment response predictor identification, as opposed to assessing specific genotype-drug interactions.

Licinio and colleagues utilize an alternative strategy to GWAS—the candidate gene approach—to detect a modest relationship between a variant in the BDNF gene and antidepressant response. The candidate gene strategy provides a hypothesis-driven approach that may limit the amount of tests conducted, and thus a smaller, more homogenous sample can be utilized. The lack of comprehensive assessment of the entire genome, as well as concerns about how the “candidacy” of each gene is assessed merit caution, but candidate gene strategies remain a particularly powerful approach in pharmacogenetic studies, where it can be more reasonably argued that we have prior knowledge of some aspects of the mechanism of drug action. Replication and extension of these results will be informative, but, as noted above, it will be critical to examine the phenotypes of putative “replication” samples. The publication of these results do provide an opportunity for maximally informative replication studies to be designed going forward.

In conclusion, these two studies suggest that pharmacogenetic studies offer intriguing insights into psychotropic drug efficacy. Although neither result may be “ready for the clinic” at this point, continued progress at both the phenotypic and genotypic levels, as well as the increased attention to pharmacogenetics at the industry level, are encouraging signs. Studies such as these are important milestones along the road to the development of personalized medicine in psychiatry.

View all comments by Anil Malhotra

Related News: Psychiatry Aims to Get Personal in Studies of Genes, Drug Efficacy

Comment by:  Alessandro SerrettiAntonio Drago
Submitted 6 July 2009
Posted 6 July 2009

Licinio and colleagues undertook a thorough analysis of the BDNF gene, in a medium-large sample of 264 controls and 272 majorly depressed patients. The study finds its rationale on the assumption that the identification of novel rare variants located in candidate genes may provide some molecular and disease-related breakthroughs. Licinio and colleagues report convincing evidence that rs12273539 and rs61888800 may be modulators of the risk for major depression disorder and of the response to antidepressant treatment, respectively. Moreover, two haplotypes in different blocks were found to be associated with the risk of being diagnosed with a major depressive disorder. The robustness of the applied statistics and the high-ranked technical methodologies which characterize the paper make it worth being taking into consideration for further analysis and independent confirmatory investigations.

Nonetheless, it is common opinion that even the most intensive methodological analysis of a single gene may not sort out the key disruptions that likely lead to major depressive disorder, in that there is good reason to think that a multigenic background characterizes psychiatric disorders. It is rational to think that a set of variations with functional roles, associated with coding or regulatory events, promotes a shift from homeostasis to allostasis, which may or may not result in a frank disease until the allostatic load is maintained. In other words, controls may bear some variations in the BDNF gene which are shared by cases and which influence risk of a depressive disorder only in the case of specific combinations with other variations not necessarily located in the gene under investigation. Moreover, gene X environment analysis may grant some new insight in this sample of patients. These elements, epistasis and gene X environment interaction, provoke a noise effect which seemed, nevertheless, to be overridden by the strength of association, Licinio and colleagues report, suggesting that rs12273539 and rs61888800 should be taken into serious consideration when designing new genetic association studies on major depressive disorder.

Perlis and colleagues report on a GWAS focused on the pharmacogenomics of the outcome of lithium treatment in a large sample of bipolar I and bipolar II patients. Unfortunately, none of the 1.4 million SNPs analyzed reached the genomewide significant threshold (10-8), and the authors could only report a suggestive association with a borderline significance (10-7) for rs10795189. This is not conceptually surprising as such a high threshold would imply a large effect of the gene, which is unlikely, and more probably true associations are hidden within results with much lower significance, but located in key mechanisms. In fact, the authors report that associations were detected between regions coding for a subunit of the ligand-gated ionotropic glutamate receptor, GluR2/GLURB, Syndecan-2 (SDC2) which codes for a cell-surface proteoglycan, synaptic vesicle, glycoprotein- 2B which plays a role in the hippocampus, and the human homologue of Drosophila odd Oz.

Two relevant points arise from this result: 1) GWAS still do not yield the expected results in the field of psychiatry, continuing to provide starting points more than solutions to the open questions in psychiatry; 2) lithium response is under the control of multiple genes, as the authors conclude, as well as under the influence of a set of environmental and psychological variables which should be taken in a more detailed account in the next investigations. Interestingly, the most common proteins which have been thought to be associated with lithium’s efficacy did not show themselves in this paper: inositol monophosphatase (IMPase [isoenzymes 1 and 2]); inositol polyphosphate 1-phosphatase (IPPase); bisphosphate nucleotidase (BPNase); fructose 1,6-bisphosphastase (FBPase); phosphoglucomutase (PGM) and glycogen synthase kinase-3 (GSK-3β) are inhibited by lithium at therapeutic concentration but their variation was not labeled as significant in this research. This stresses the prime relevance of GWAS in identifying new molecular paths associated with drug dynamics which could be quite far away from the theories that have been so far been built in vitro and in animal models: this kind of falsification strategy is of major help in stimulating new and hopefully more informative directions of investigation.

View all comments by Alessandro Serretti
View all comments by Antonio Drago