Heterogeneity of clinical response to antipsychotics

The diagnostic classification of schizophrenia relies on the presence of specific symptoms and the course of the illness, without there being any diagnostically distinguishing molecular or genetic biomarkers to aid the diagnostic process. The phenotypical presentation of schizophrenia is highly heterogeneous, a characteristic that is reflected in the variability of response to antipsychotic medication in patients with the disorder. Approximately 50–70% of patients with a first episode of schizophrenia will respond to antipsychotic medication, with this response rate falling to 20% for those who require a second trial (Reference Agid, Arenovich and SajeevAgid 2011).

Although antipsychotics are the mainstay of treatment for schizophrenia, about a third of patients do not respond to them and are described as having treatment-resistant schizophrenia. Further, non-adherence to antipsychotic treatment is high and up to 75% of patients are not adhering to their antipsychotic medication regimen 2 years after discharge from hospital (Reference Leucht and HeresLeucht 2006). In large part this is due to intolerable adverse effects, leading to the discontinuation of antipsychotics.

The high cost of antipsychotic treatment failure

Antipsychotic treatment failure and intolerability comes with a high clinical and economic cost. The proportion of patients who recover from schizophrenia is far from impressive, with a recent systematic review estimating that only one in seven patients attain a functional recovery (Reference Jääskeläinen, Juola and HirvonenJääskeläinen 2013), and only 20–30% of people with schizophrenia have a complete remission of symptoms (Reference Fenton and McGlashanFenton 1991).

The cost of care for schizophrenia was estimated to be £6.7 billion in the UK alone in 2004–2005 (Reference Mangalore and KnappMangalore 2007), and the total societal cost in England is estimated at £11.8 billion per year, with a cost to the public sector of £7.2 billion (Reference Andrew, McCrone and ParsonageAndrew 2012). The cost of care for individuals with treatment-refractory schizophrenia is very high, with annual costs for patients with treatment-resistant schizophrenia being 3- to 11-fold higher than for patients with treatment-responsive schizophrenia (which are $15 500–$22 300 per annum in the USA) (Reference Kennedy, Altar and TaylorKennedy 2014). These findings reinforce the need to find better ways to predict antipsychotic treatment response and tolerability.

This heterogeneity in treatment response and tolerability in schizophrenia has led to emerging attempts to stratify the patient population by treatment response. In the rest of this review, we will focus on the general application of genomic and other biomarkers for the prediction of treatment response in schizophrenia and on obstacles to personalised medicine for the disorder.

Pharmacogenetic testing and clozapine response

Pharmacogenetic studies have investigated the importance of serotonergic and dopaminergic targeting for the antipsychotic effect of clozapine, but the results are conflicting and poorly replicated (Reference Zhang and MalhotraZhang 2011). No single polymorphism can be used for the prediction of clozapine response. The failed attempts to implement findings of candidate gene studies in predicting clozapine response is emblematic of the difficulties in the clinical utilisation of this testing in psychiatry. The lack of translational impact of candidate gene testing in predicting the efficacy of clozapine and other antipsychotics has been a hallmark of studies in this area and has led to more recent efforts to utilise GWAS and polygenic risk scores to provide more robust genetic markers.

Other putative biomarkers

Medication hypersensitivity

Although there are few examples of the use of pharmacogenetic testing to maximise treatment response in clinical practice, pharmacogenetics is more widely used in the prediction of hypersensitivity to certain medications such as carbamazepine. In people of South East Asian ethnicity, positivity for the HLA-B*1502 allele conferred an odds ratio of 2504 for developing carbamazepine-induced Stevens–Johnson syndrome/toxic epidermal necrolysis (Reference Ferrell and McLeodFerrell 2008). Similarly, screening for HLA-B*5701 reduces the risk of a hypersensitivity reaction to abacavir in HIV treatment, with positivity for the allele providing 100% specificity for the development of an hypersensitivity reaction (Reference Mallal, Phillips and CarosiMallal 2008). Both are examples of how a pharmacogenetic test can be used to avert a specific medication toxic effect.

Antipsychotic adverse effects

The prediction and avoidance of adverse events is an important goal in the use of antipsychotic medication. Two of the most widely studied adverse effects in pharmacogenetics are tardive dyskinesia and weight gain.

Probably the most robust evidence for a genetic finding associated with side-effects is that for antipsychotic-induced weight gain, with the 759T/C polymorphism in the promoter region of the HTR2C gene (rs3813929) associated with a greater than twofold increased risk of weight gain (Reference Arranz, Rivera and MunroArranz 2011). This effect is especially marked with olanzapine and clozapine, both of which have strong affinity to the 5-HT2C receptor.

With regard to tardive dyskinesia, despite several studies of dopamine receptor polymorphisms, the results have been inconsistent and there remains no pharmacogenetic test for the development of antipsychotic-induced tardive dyskinesia (Reference Ozomaro, Wahlestedt and NemeroffOzomaro 2013).

Clozapine and agranulocytosis

Clozapine is underprescribed and its use is limited by patients’ and clinicians’ fears about the emergence of life-threatening adverse events, in particular agranulocytosis. The mechanism of this idiosyncratic drug reaction is yet to be elucidated, although it is certainly multifactorial, with some evidence that genetic variance increases individuals’ susceptibility to clozapine-induced agranulocytosis. It remains an area which has not been extensively studied and, owing to the relatively low incidence of clozapine-induced agranulocytosis, it is dominated by small-sample studies.

The most replicated genetic evidence is for variation in the human leucocyte antigen (HLA) system. This includes several genetic variants involved in the HLA-DQB1 locus, including associations at two HLA amino acids (a single amino acid at HLA-DQB1(126Q) and an amino acid change in the extracellular binding pocket of HLA-B(158T)) (Reference Goldstein, Jarskog and HilliardGoldstein 2014). Previously, a case–control cohort found that the odds of developing clozapine-induced agranulocytosis were 16.9 times higher in patients with the sequence variant 6672G>C on the HLA-DQB1 gene (Reference Athanasiou, Dettling and CascorbiAthanasiou 2011). This single nucleotide polymorphism has been incorporated into commercially available tests, with test sensitivity and specificity calculated to be 21.5% and 98.4% respectively, indicating that about one in five patients at risk of clozapine-induced agranulocytosis would be identified by the test (Reference Athanasiou, Dettling and CascorbiAthanasiou 2011). Although the HLA-DQB1 locus may be implicated in the development of clozapine-induced agranulocytosis, for the test to have clinical utility for this adverse event, which has a prevalence of 0.8%, it requires high sensitivity as well as specificity. The low sensitivity of the test has been a limiting factor in its application in clinical settings; and it has not been a commercial success (Reference Chowdhury, Remington and KennedyChowdhury 2011; Reference Verbelen, Collier and CohenVerbelen 2015).

Complexity and heterogeneity of the schizophrenia phenotype

Thus far, there has been little success in developing personalised medicine for schizophrenia. The reasons for this are manifold; not least that schizophrenia is a highly complex disorder influenced by multiple combinations of genes and molecular variances acting in the midst of even less clearly definable social and environmental factors, thus limiting the likelihood that a single biological risk factor will map onto a diagnosis of a psychotic disorder.

Lack of a mechanistic understanding of antipsychotics

The medication response phenotypes in psychiatry are based on clinical observation and not on the physiological processes or biomarkers that can be utilised in other medical specialties such as oncology and cardiovascular medicine. Our limited understanding of the mechanism of action of antipsychotics (compared with chemotherapeutic agents in oncology, for example), limits our ability to generate mechanistic hypotheses as to which biomarkers might be relevant to treatment response.

Clinical guidelines and interpretation

It is noteworthy that there are 121 known pharmacogenetic markers for medications currently listed for prescription by the US Food and Drug Administration, with a sizable minority (n = 26) for psychiatric medications (Reference HamiltonHamilton 2015). However, specific and clear guidelines for the clinical use of these pharmacogenetic markers are lacking. This highlights a recurring problem, not only in psychiatry, but across other medical specialties as well.

The most common psychotropic medication markers relate to CYP2D6 enzymes. This particular focus for pharmacogenetic research on CYP2D6 genotypes is not unexpected, given that this enzyme metabolises a number of antipsychotics, with most of the research focused on the relationship with risperidone use. Genotyping for CYP enzyme deficiencies could potentially offer a relatively simple solution for optimising antipsychotic medication dosing, although this has not been consistently borne out in practice and genetic testing for CYP polymorphisms is not currently in routine use for guiding antipsychotic choice or dosage. The predictive value of currently available tests is not sufficient to allow the integration of pharmacogenetic testing into clinical guidelines.

Clinician and public attitudes

Although clinicians and patients are generally positively disposed to the concept of personalised medicine, the introduction of personalised approaches will require a profound change in prescribing practice, with treatment decisions taken out of the hands of clinicians and patients and devolved to algorithms based on genetic tests or other biomarkers. This represents a growing challenge for medical education, if future generations of doctors are to be prepared for advances in the field (Reference SlaughterSlaughter 2012).

Communication and interpretation

The communication of pharmacogenetic test results is a critical step towards their widespread utilisation and this has been highlighted across medical specialties. In the future, the focus will shift from the costs of genotyping to the interpretation and production of reports for clinical use. The need to organise and interpret complex pharmacogenetic testing will rely on the utilisation of more highly integrated electronic patient records, to ensure that this information is available for all clinicians during any future clinical encounter.

Positive pharmacogenetic test results will have lifelong implications for a patient's response to, or tolerance of, antipsychotic medication. However, in a fragmented healthcare system, there is a risk that access to pharmacogenetic test information will be limited and that it will be lost to clinicians as the patient moves from one healthcare setting to another.

‘Orphan diseases’

A wider ethical question surrounds the potential for the creation of so-called ‘orphan diseases’ or, in patients with schizophrenia, a subset of patients with ‘Cinderella psychoses’. The implementation of reliable pharmacogenetic testing will inevitably result in the subdivision of patients into smaller subgroups. This would bring with it the possibility that some of these subgroups might be considered less profitable by the pharmaceutical industry, and thus stand a reduced chance of having treatments developed for them, particularly where markers are found that define patients with a low probability of treatment response. Further, there may be potential difficulties in finding adequate numbers of specific patient groups in order to identify variance in individual biomarker profiles. An example of this is the identification of patients at risk for clozapine-induced agranulocytosis, which due to its low incidence and unpredictable occurrence makes it difficult to assemble statistically adequate samples for prospective study. Patients with less profitable genotypes are therefore at risk of becoming therapeutic orphans.

Health economics

The cost-effectiveness of any personalised medication approach will depend on the prevalence of test-positive individuals, the accuracy and the cost of conducting the test, the predictive validity of the test and the cost of any newly developed medication as a result of the test (Reference Flowers and VeenstraFlowers 2004).

There is little research assessing the economic evaluation of the costs and outcomes of the use of pharmacogenetic tests in schizophrenia. We located one recent evaluation of the potential use of a pharmacogenetic test assessing CYP450 polymorphisms prior to treatment with risperidone (Reference Rejon-Parrilla, Nuijten and RedekopRejon-Parrilla 2014). In this study, pharmacogenetic testing prior to risperidone treatment was used to guide dosing and was compared with the usual ‘trial and error’ dosing scheme. The patient stratification strategy by pharmacokinetic testing improved health compared with traditional dosing schemes at a cost of £2059 per patient. This study's results are promising and indicate that the pharmacogenetic test for risperidone in schizophrenia could meet the requirements of cost-effectiveness set by the National Institute for Health and Care Excellence (NICE). One counter-argument to this approach is that it may be more cost-effective simply to treat with antipsychotics for which common CYP450 variation has little or no effect (e.g. amisulpride).

The stratification of patients with psychotic illness into more refined subsets may lead to increased efficiency of medication development, but with a reduction in the size of the total schizophrenia population. This may lead to the development of multiple subsets of patients and niche products, which will risk segmenting the market and increasing drug development costs (Reference EversEvers 2009), resulting in more expensive medications for small patient groups, which will not be economically viable products. This may in turn act as a barrier to future medication development by pharmaceutical companies.