OPtimization of Treatment and Management of Schizophrenia in Europe (OPTiMiSE)

Project Information

OPTIMISE

Grant agreement ID: 242114

Project closed

Start date 1 February 2010

End date 31 July 2016

Funded under

• Specific Programme "Cooperation": Health

Total cost € 14 900 190,31

EU contribution € 11 187 685,42

Coordinated by

UNIVERSITAIR MEDISCH CENTRUM UTRECHT
Netherlands

Final Report Summary - OPTIMISE (OPtimization of Treatment and Management of Schizophrenia in Europe (OPTiMiSE))

Executive Summary:
Despite nearly fifty years of pharmacological and psychosocial research, the overall prognosis of schizophrenia has improved only marginally. While the antipsychotic efficacy of most antipsychotics is generally uncontested, their overall functional impact has been modest. OPTiMiSE (OPtimization of Treatment and Management of Schizophrenia in Europe) has focused on two main goals: a) optimizing current treatment through a series of clinical studies which push the boundaries of existing treatments in their application and implementation and b) exploring novel therapeutic options through an integrated series of experimental studies based on novel strategies, and novel imaging and genomic technologies. The project has been divided into multiple Work Packages, most of which have produced (preliminary) results. Others, mainly those focusing on metabolomics and proteomics, are expected to do so before the end of 2016.

First, a treatment algorithm was tested, showing that amisulpride appears to be a good option for initiating pharmacotherapy in first episode patients. In contrast to expectations, there was no clear advantage regarding effectiveness for non-remitters (after 4-week amisulpride treatment) to either switch to another antipsychotic or remain on amisulpride treatment for another 6 weeks. The switch to clozapine in patients who had not reached remission after the first 2 study phases resulted in a low percentage of remission. Nevertheless, patients who did not meet the strict remission criteria often did show a substantial reduction in psychotic symptoms.

Second, from various angles, biological markers were compared between remitters and non-remitters. Even though not all analyses were completed yet, we did identify various differences; a smaller cortical thickness was found in various areas for non-remitters compared to remitters, however none of these findings survived correction for multiple comparisons. Following first genetic analyses, 26 genes were differentially expressed between remitters and non-remitters. Data from preliminary analyses suggest that S100B levels are correlated with the severity of PANSS positive symptoms at baseline, but that there is no simple relationship between S100B levels and antipsychotic response. No significant differences were found in brain glutamate measures at baseline between remitters and non-remitters.

Third, we tested whether it would be possible to predict who will respond and who will not, based on various biological markers. When utilizing the neuroimaging data (WP1), the preliminary results show that is possible to predict whether a subject will respond with an accuracy that is greater than chance level. This ability is greater when using white matter data than grey matter data. While glutamate levels at baseline could not predict whether a patient will reach remission, glutamate levels at baseline may be associated with subsequent symptom severity and predict the extent to which symptoms may be expected to decline. Serum cytokines were found to be potentially relevant biomarkers for predicting whether or not FEP patients will meet remission criteria when treated with amisulpride.

Once all analyses are completed, machine learning methods will be applied to test whether the combination of various demographic data, clinical assessments, neuroimaging and blood marker data can predict response and consequently aid clinicians in making informed decisions on the treatment of first episode patients.

For two Work Packages, primary results are still pending: the data collection for the Psychosocial Intervention component (Work Package 3) is still ongoing and the first results are expected in October 2016. Unfortunately, the ambitious objective of Work Package 4, demonstrating efficacy of cannabidiol in the treatment of first episode psychosis, was not met due to difficult circumstances in the trial management (e.g. development and manufacturing of study medication) as well as difficulties in recruitment due to the extensive safety measures required for placebo-controlled trials. This trial is expected to continue in collaboration with a new funding source.

Project Context and Objectives:
While we have had effective antipsychotic treatments for nearly fifty years, the application and implementation of these treatments is far from optimal. Many of the elementary questions in the treatment of schizophrenia have remained unanswered. When a psychiatrist is faced with a new patient with schizophrenia he will no doubt use an antipsychotic to start treatment; however, he has little guidance on some very simple and fundamental questions. For instance, is there a rational basis for choosing the first antipsychotic? Can I predict how well the patient will do? When the patient fails to respond to the first antipsychotic, how long do I wait? Do I continue for some more time or do I switch to another antipsychotic? If so, which one? Fortunately, first episode patients do often respond reasonably well; the main challenge then becomes how to keep them well. The single best
predictor of continued wellness for the patient is compliance with treatment. And while every psychiatrist knows this to be the case, there are few, if any, simple, effective and widely applicable manoeuvres at their disposal to increase compliance. Answers to these questions are critical to optimising the treatment of schizophrenia; hence this is the first focus of OPTiMiSE.

We have addressed each of these questions in an integrative study including almost 500 first-episode schizophrenia patients with minimal prior exposure to antipsychotic treatment who will be followed for one year. We expect the answers to these questions to have immediate and practical benefits for patients, carers and health care professionals and for the health care delivery in Europe.

Objectives for component A - optimizing current treatment:
1) To use MRI to optimize treatment outcome through the exclusion of patients whose psychotic symptoms are not due to the disorder but to underlying 'organic' pathology; antipsychotic treatment in these patients is inappropriate and may delay the initiation of potentially life-saving medical
intervention.
2) To provide a rational basis for antipsychotic choices in the treatment of first episode schizophrenia or schizophreniform disorder, which is threefold; i) determine whether amisulpride is a sensible first choice; ii) study the optimal treatment choice for non-responders (continue treatment versus switch to antipsychotic with other mechanism of action); iii) in line with current treatment guidelines but contrary to clinical reality, non-responders will be switched to clozapine to determine whether clozapine should be used early in the treatment course, instead of a last resort.
3) To improve functional outcome and reduce drug discontinuation in symptomatically remitted first episode patients by means of widely applicable psychosocial interventions, partly IT-enabled.

Objectives for component B - exploring novel therapeutic options:
1) To definitively assess the predictive value of MRI data for treatment response, through previously untreated patients who are scanned prior to the administration of a standard treatment, with their response assessed prospectively.
2) To test whether glutamatergic markers, assessed through MRS imaging technologies, predict response to first and second line treatments.
3) To explore the potential of cannabidiol CBD, a modulator of endocannabinoid functioning, as an alternative to D2 based antipsychotics.
4) To examine whether an empirical combination of pharmacogenetic, proteomics- and metabolomics markers can provide a clinically valuable predictive value.

Project Results:
Work Package 1: Structural MRI as a predictor of treatment response
Objective # 1: To define the nature and prevalence of ’organic’ pathology in patients presenting with a first episode schizophreniform psychosis
Schizophrenia and psychoses in general, present with a combination of symptoms, such as auditory hallucinations or delusions, which do not point to the specific origin of these symptoms. While in many cases a specific underlying organic disorder is not identifiable (functional psychosis), it has long been acknowledged that in a subset of psychotic patients, symptoms are attributable to a medical or neurological disease, such as sarcoidosis, porphyria, Cushing’s disease, acquired immune deficiency syndrome, carcinoid tumor, toxic thyroid nodules, Wilson’s disease, neurosyphilis, Huntington’s disease, or multiple sclerosis. These are the so called ‘secondary’ or organic psychoses, to indicate they are due to a general medical condition. The exact percentage of organic psychosis is not well known. However, it is important to identify these medical or neurological causes of psychosis early, as they require urgent treatment of the primary disease instead of treatment with antipsychotic drugs. Unfortunately, it is not always possible to differentiate ‘primary’ and ‘secondary’ psychosis just on the basis of clinical presentation. For this reason, some countries (for example Germany and Denmark) now use brain imaging to rule out secondary psychosis as a routine screening tool for all patients presenting with psychotic symptoms. However, whether this assessment should be implemented in clinical practice remains subject of debate in other countries. To make definitive recommendations as to whether or not MRI should be considered as part of routine screening protocols, it is important to establish what is the proportion of patients presenting with clinically relevant neuropathology in brain scans, since neuropathological findings identified with magnetic resonance imaging may be rare and may not outweigh the associated burden and costs (for example: 0.2% prevalence of brain tumours in healthy subjects).

Over the course of the study, a total of 258 magnetic resonance imaging (MRI) scans were obtained from patients recruited at 8 of the participating centres. Of these, 203 were acquired at baseline and 55 at the 4-week follow up. In addition, 163 healthy controls were scanned. With such a large group of subjects divided over multiple centers, first it is essential to make sure the MRI scans scan in a similar way and to test for variability in scanners between the participating sites; this way one can ensure that grey and white matter boundaries remain properly distinguishable, so that the segmentation algorithms are able to perform optimally during data-analyses. Therefore, for quality control purposes, a total of 80 MRI scans were acquired from ‘quality control’ subjects who were scanned several times either within each centre or across multiple centres. Additionally, phantom measures were obtained. The phantom data showed good inter and intra scanner stability between participating centers. This provides confidence that the human data acquired across the different centres can be reliably merged for the study analyses.

A total of 203 MRI from the patients recruited in the study were of a quality suitable for examination. The most common findings at baseline included: non-specific white matter T2-weighted hyperintensities (n=48); cavum septi pelludici (n=34); arachnoid cysts (n=9); brain volume loss/atrophy (n=4); connatal cysts (n=1); cavum septi pelludici and cavum vergae (n=1); mega cisterna magna (n=1); pineal cyst (n=1). In terms of organic pathology, we found a total of n=3 neoplasms. There were a further n=4 cases that required further investigation and were referred accordingly: 3 cases required further diagnostic MRI scanning with appropriate protocols so that a definitive diagnosis can be established (n=1 vermian cyst to be characterised with diagnostic MRI head with contrast; n=1 likely cortical dysplasia to be characterized with an epilepsy MRI protocol; n=1 cystic lesion of unclear origin requiring diagnostic MRI scan including with contrast); and 1 case (n=1 silent sinus syndrome) that required follow up by a Nose-Ear-Throat specialist.

We identified a total of n=7 patients with findings that required a notification to the responsible clinicians for further follow up; n=3 were definitive evidence of organic pathology (neoplasms). This evidence suggests that the proportion of organic pathology of clinical significance (i.e. neoplasms) relevant for psychosis is rather small, only around 0.01%. This rate of organic pathology would be too low to justify the inclusion of MRI screening for all patients presenting with a first episode of psychosis. However, the proportion of white matter hyperintensities and brain atrophy, which are of uncertain but potential clinical significance, accounted to 25%. If these were to be considered of clinical significance, MRI would then represent a potentially useful tool in the routine initial clinical screening of psychosis onset.

Objective # 2: To determine the extent to which MRI measures at first presentation predict the therapeutic response to subsequent antipsychotic treatment
The most pressing challenge in individuals experiencing the first episode of schizophrenia is how to predict who will respond to an antipsychotic with a specific pharmacological profile, so that symptom remission is achieved quickly. Magnetic Resonance Imaging (MRI) measures are promising biological markers for treatment outcome. In fact, schizophrenia is associated with robust reductions in regional frontal and temporal grey matter volumes, and enlargement of the ventricles, and these have been associated with more severe symptom profiles as well as poorer outcome. Furthermore, previous studies suggest that patients who do not respond to treatment compared to those who do, have thinner cortical thickness in the frontal, temporal and parietal cortices. Importantly, these research findings represent mean differences between groups of patients who differ in terms of their antipsychotic response. However, these research findings need to be translated into tools that could allow clinicians to stratify individual patients according to their likely future therapeutic response, to have clinical utility.

OPTiMiSE has given us the opportunity to study a large, clinical sample of individuals homogenous in terms of illness stage and treatment. In OPTiMiSE we investigated: 1) differences in spatial distribution and volumes of grey matter between patients who responded and patients who did not respond to treatment at 4 weeks (primary outcome); and 2) the extent to which grey matter volumes at presentation predicted response to antipsychotic treatment (secondary outcome).

A total of 203 patients had a baseline scan. However, some scans could not be included in the grey matter analyses. Reasons for exclusion were: MRI scan or clinical information not available (n=11); structural abnormality in the brain (n=8); motion artifact (n=22); poor segmentation (n=1). Thus, scans from a total of n=161 patients could be included in the main grey matter volume and treatment response prediction analyses.

The current results are still preliminary, as the analyses on the locked dataset are currently ongoing. Group comparisons revealed clusters of differences between Remitters and Non-Remitters in both hemispheres. More precisely, for the left hemisphere, Non-Remitters showed smaller cortical thickness than Remitters in the superior-parietal, postcentral, precentral, and supra-marginal regions and higher thickness in superior temporal and posterior cingulate regions. However, none of these cluster survived correction for multiple comparisons.

In the right hemisphere 6 clusters were identified. In particular, Remitters showed larger thickness than Non-Remitters in the superior-parietal and inferior-temporal areas, and smaller thickness in middle-temporal, para-hippocampal, and superior-temporal areas. However, none of these cluster survived correction for multiple comparisons. Results are shown in figures 1-4.

Volume of grey matter
There was no significant difference between Remitters and Non-Remitters in terms of total intracranial volume, total grey matter volume, total white matter volume and total cerebrospinal fluid volume.

We found a distributed pattern of smaller grey matter volumes in Non-Remitters as compared to Remitters. This was particularly evident in frontal, parietal and visual cortical regions. However, none of these cluster survived correction for multiple comparisons. Results are shown in figure 5.

Treatment response prediction
We conducted two sets of analyses: first, we used the whole dataset (including the 161 scans from all centres) in the training stage; and second, we repeated the analysis but using only data from single centres in the training phase (thusby using smaller sample sizes but improving homogeneity by including only scans acquired in one centre). This was done for all the countries that had a reasonable number of subjects and where the two groups (Remitters and Non-Remitters) were balanced in size. These were: Denmark, Israel, Spain and the UK. Below (Table 3), we report the results for the whole group of 161 subjects, and also for the individual centres. The P-values are given for an improvement (higher value than) in the equivalent null distribution statistic; where none is given, the experiment gave worse results than the permutation test. Results are shown in table 1.

Conclusions Work Package 1
Spatial distribution of grey matter
This preliminary analysis indicates the presence of differences in cortical thickness between Remitters and Non-Remitters. Future work will enable further investigation of subgroups within the patient group that can explain differences in treatment response, by combining neuroimaging information with clinical assessments and other biological markers. Longitudinal clinical assessments will also allow us to assess progression of the symptomatology over time and their association with the observed brain alterations.

Volume of grey matter
Further data-analyses will be conducted to include region of interest analyses, and with a less conservative approach in terms of scan exclusion.

Treatment response prediction
The results primarily show that is possible to predict whether a subject will respond with an accuracy that is greater than chance level. This ability is greater when using white matter data than grey matter data, although the permutation test showed that the accuracy was not statistically significant. When examining subjects within a single country, there was a wide variation in accuracy. Thus, simply examining subjects from a single centre in itself is not sufficient to improve classification. On the other side, the overall accuracy of 57% is smaller than the accuracy achieved from some of the single country datasets (see UK for example). This suggests that, while pooling results across multiple centres increases the volume of training data, it may introduce extra inhomogeneity into the data (not related to the Remitter/non-Remitter condition), which may outweigh the effects of greater numbers.

Caution must be taken when interpreting these preliminary findings, as most countries included a substantially greater proportion of Remitters than Non-Remitters in their datasets. This means that a poor classifier can still achieve a deceptively high accuracy by predicting the majority class for most subjects. Nevertheless, the findings also suggest that it is possible to achieve both good sensitivity and specificity. This can be seen for example in the white matter predictions, where among the Israeli patients sensitivity, specificity and overall accuracy were all well above 50% by a statistically significant margin (although this significance would not survive rigorous control for multiple comparisons). The predictions for subjects scanned in Denmark are also well balanced, as there was an almost equal numbers of Remitters and Non-Remitters in those subjects.

Future work may benefit from greater numbers of subjects, but it must be accompanied by an effort to ensure the scanner and scanning protocols as well as diagnostic procedures in all countries as similar as possible to maximise the benefits of integrating large numbers of subjects from multiple centres. Statistical methods to reduce the confounding effects of multiple centres after the fact would also be of interest.

Work Package 2: Switching
Very few prospective, sequential studies are available that could guide decisions which have to be made in every day clinical routine. Some of the simplest questions of clinicians remain unanswered. For example: If the first antipsychotic used has not worked, is switching to another drug effective? Or should we perhaps increase the dose? And when should we start clozapine, the most efficacious drug? These questions are most urgent for patients with a first episode of schizophrenia: optimal treatment in this early phase is of crucial importance, as swift restoration of social and professional functioning improves long-term outcome. First episode patients on average respond better and faster to antipsychotic drugs, need lower doses and have an overall better prognosis than chronic patients. The first episode is therefore a critical junction in the lives of people with schizophrenia, where optimal treatment could positively influence the long-term course.

In current literature, consensus has been reached that clozapine is the most efficacious antipsychotic drug, but it is associated with severe side-effects, especially potentially fatal agranulocytosis requiring weekly blood monitoring. Most treatment guidelines suggest that clozapine should be started after at least two other antipsychotic drugs, which were given in sufficient doses and for a sufficient duration, have failed. According to current recommendations, a patient who fails two trials of 6 weeks with an antipsychotic should be offered clozapine, which would imply that a first episode patient should be offered clozapine within 12 weeks of start of treatment. However, this is hardly ever the case: standard treatment evidence shows that the average patient being initiated on clozapine has often been psychotic for nearly 10- 12 years. Results published in 2007 suggest that clozapine can be systematically applied as the third line treatment within the first 6 months, and show dramatic benefits for those who had failed two other atypical medications (Agid et al., 2007).

OPTiMiSE is one of the rare studies to test a treatment algorithm. Almost 500 participants with a first episode of schizophrenia, schizophreniform or schizoaffective disorder were treated according to a 3-phase pharmacological algorithm shown in Flowchart 1: after a 4-week open label amisulpride treatment period, non-remitters were randomized to either switch to another antipsychotic versus continuation of amisulpride for 6 weeks, in a double blind fashion. After these additional 6 weeks, non-remitters are switched to clozapine for another 12 weeks. The study design in shown in flowchart 1.

479 patients with a first episode of psychosis were included in a total of 22 institutes located in Europe and Israel. For the current analyses, patients recruited at one of the Romanian centers were excluded due to a suspicion on scientific fraud. In addition, for 12 patients, their eligibility is still under review; therefore they were not included in the current analyses. In total, 459 patients were included in the current analyses. Mean age at inclusion was 25.6 years, 30.3 % of the patient sample was female. The median duration of illness was 4 months and none of the participants had been using antipsychotic medication for more than 3 weeks. All patients were treated on a voluntarily basis. The patient sample was moderately ill at baseline, with a mean score on the Positive And Negative Syndrome Scale (PANSS) of 78 (sd 19).

Phase 1
459 patients started on 200-800 mg/day amisulpride treatment, with a target dose of 400 mg/day. A total of 80 patients dropped out after the medication was initiated, whereas 379 patients reached the end of the first 4-week treatment phase. Out of the 459 patients who were initiated on amisulpride, 55% met remission criteria, defined as the eight remission items of the PANSS. Side effects were mild, with a mean weight increase of 2.6 kg (sd 4.2).

Phase 2
At the end of phase 1, 129 patients did not meet remission criteria. These patients were asked to continue in phase 2. A total of 30 patients decided not to continue into phase 2. Out of the 99 patients who initiated the double blind treatment, 6 patients were excluded from the analyses due to the issues described above. 20 patients dropped out, whereas 73 patients completed the 6-week treatment phase. At the time of writing the current report, treatment assignments are still blinded, as the database has not been officially locked yet. Therefore, the mean dose of amisulpride and olanzapine cannot be reported yet.

Out of the 93 patients who started double blind treatment and were included in the current analyses, 44% met remission criteria at the end of phase 2. The remission rate did not differ significantly between the two treatment arms, with 43.2% for treatment A and 44.4% for treatment B (p=0.92). The mean PANSS score did not differ significantly between treatment arm A (68.3 sd 19.6) and B (67.8 sd 17.5). In addition, the drop-out rate was not significantly different between treatments, with 22.4% in treatment arm A and 17.8% in treatment arm B (p=0.57). These findings suggest that, when a patient does not improve sufficiently after 4 weeks of amisulpride treatment, there is no clear advantage to switch to another antipsychotic with a different receptor binding profile over continuing the initial antipsychotic treatment for another 6 weeks. The slightly lower than planned number of participants (479 instead of the targeted 500) did not have a negative influence on the statistical power: the differences between groups would not have reached statistical significance in in case 500 patients would have been recruited.

Phase 3
At the end of phase 2, 35 out of the 77 completers met remission criteria. Out of the 42 patients who did not meet remission criteria, 29 patients continued into phase 3 and initiated clozapine treatment. During this treatment phase, 11 patients dropped out, mainly because they did not want to participate in clinical research anymore (n=5) and due to a lack of efficacy (n=4). Of the 18 patients who completed the 12 week clozapine treatment, only 5 patients met remission criteria, translating into a remission rate of only 17% (based on the group of patients who initiated clozapine treatment). Patients gained a mean weight of 5.2 kg (sd 5.4).

Even though a higher remission rate was expected within this group, based on the earlier report by Agid and colleagues (2007), the results are not as discouraging as they may seem at first glance. The remission criteria as utilized in this clinical trial, which are widely used in this scientific field, are very strict; if a patient does not meet remission criteria, this does not mean that he or she did not respond (well) to the medication. Indeed, when looking into the improvement in psychotic symptoms, expressed as scores on the Positive And Negative Syndrome Scale (PANSS), the patients who did not meet remission criteria at the end of phase 3 did respond well to clozapine. They simply did not improve enough to meet remission criteria. In most cases, the negative symptoms minimally improved in non-remitters, whereas the positive and general symptoms were reduced to a great extent (figure 6).

Conclusions Work Package 2
Amisulpride appears to be a good option for initiating pharmacotherapy in first episode patients, as 55% of participants reached remission criteria after 4 weeks. These findings are in line with several previous studies, among which the EUFEST trial (Kahn et al., 2008), which demonstrates the superior efficacy of amisulpride in schizophrenia treatment. However, the remission rate in phase 1 of the current study is significantly higher compared to previous trials (55% versus 40%). The underlying cause for this large difference cannot be explained at this time, but will be topic of investigation in the coming months. In contrast to expectations, there was no clear advantage regarding effectiveness for non-remitters to either switch to another antipsychotic or remain on amisulpride treatment for another 6 weeks. The switch to clozapine in patients who had not reached remission after the first 2 study phases resulted in a low percentage of remission. Nevertheless, patients who did not meet the strict remission criteria often did show a substantial reduction in psychotic symptoms.

Work Package 3: An IT-enabled psycho-educational intervention to improve treatment adherence
The patients who either met remission criteria at the end of phase 1, 2 or 3, or dropped out of the trial designed for Work Package 2, were invited to participate in the Psychosocial Intervention component. The purpose of this Work Package was to test whether a combination of interventions would increase treatment adherence in first episode patients. At baseline, patients were randomised 1:1 to treatment as usual versus a psychosocial intervention, consisting of 6 Motivational Interviewing sessions, access to a psychoeducational website and an SMS-system which aided the participants in reminding to take their medication. Recruitment closed with only a slightly lower number of participants than projected: 246 patients have had a baseline visits, whereas 264 participants were expected.

The recruitment for Work Package 2 was maximally extended until April 15, 2016, based on the 6-month extension for the project granted by the EC. As a result, recruitment into Work Package 3 continued until the final patient had completed or dropped out of Work Package 2. Due to the 1-year follow-up period, Work Package 3 data collection will continue until Summer 2017. The 3-month data cleaned and locked. At the time of writing this report, the analyses were initiated but not yet completed. Until we have the opportunity to analyse patterns of missing data (once the database is complete), it is not clear whether changes in mean scores are indicative of gradual improvement across groups of an artefact of selection bias; this is applicable for all results. Nonetheless, the mean scores of assessment of functioning and disability at baseline are in the range indicating “Some difficulty... but generally functioning well”. Mean and SD Compliance Rating Scale scores (scale range 1-7, lowest indicating poor compliance) and Sellwood compliance scale scores (range 1-4, lowest indicating >90% compliance) are indicative of substantial levels of adherence for many but a large range.

Mean scores indicate relatively low level positive and negative symptoms (positive and negative subscale ranges 7-49) and total scores consistent with minimal-mild illness severity (total scale range 30-210), though the upper 2.5% are likely to score moderately high. Notably, there is little indication of substantial changes in score over time. The attitude towards drug use and knowledge about this topic are captured through the Drug Attitudes Inventory (DAI) and Knowledge About Psychosis Inventory (KAPI). DAI mean scores (scale range -30 to +30) indicate relatively positive attitudes to antipsychotics.

Motivational Interviewing therapeutic alliance (CALPAS total), number of sessions; SMS use (number of texts, duration of use); and website use data are not yet available; all these aspects of the trial are still running. Serum antipsychotic levels are also unavailable at this stage, though they will later contribute to the assessment of adherence. The final analyses on the 3-month dataset are expected beginning of October and will be reported to the European Commission as soon as possible.

Work Package 4: Cannabidiol study
In order to optimise treatment in schizophrenia, the potential of cannabidiol (CBD), a modulator of endocannabinoid functioning was explored as an alternative to Dopamine-2 based antipsychotics. A four week, multi-centre, double blind, randomised, placebo controlled, parallel group study was designed and implemented to evaluate the efficacy and tolerability of cannabidiol versus olanzapine and placebo in alleviating the positive, negative and general symptoms of schizophrenia. Unfortunately, the study was delayed significantly due to a series of unforeseen serious issues with developing and manufacturing the study medication, ranging from a very low biological availability to a placebo tablet that was distinguishable in appearance from the active medication. In addition, enrolment was complicated by the stringent in- and exclusion criteria, which were implemented following the advice of the European Medicines Agency, to facilitate a future submission for registration in the European Union. As this concerned a placebo-controlled trial, safety measures were extensive. The remaining centres were facing serious problems in identifying eligible individuals due to the extensive safety measures that were required by the authorities for placebo controlled trials. In particular, presence of suicidal ideation in the 12 months prior to participation in the trial was an exclusion criterion, which prevented many patients from participation. In addition, crystal meth use, which is also an exclusion criterion for the trial, increased significantly in schizophrenia patients during the last three years, in particular in areas of easy availability of the drug (Saxonia-Anhalt and Bavaria in Germany).

At the end of the project, 16 patients were included in the trial, 12 of which were randomised. As these numbers fall short of the originally planned 150 participants, and are much too low to perform any worthwhile statistical analyses, the blind was kept intact to ensure a continuation of the trial once a new funding source is secured. The Work Package leader is currently negotiating with a Canadian pharmaceutical company that is interested in collaborating on this project.

Work Package 5: Biological predictors
One of the major shortcomings in the current treatment of schizophrenia is that we have no valid criteria in clinical practice to decide which form of treatment should be chosen first. The identification of blood based biological markers of drug response with a good sensitivity and specificity would enable the physician to use these tests prior to choosing the antipsychotic treatment and therefore help the practitioner in his daily clinical practice. Secondly, the identification of these markers will help to identify new and more specific pharmacological targets as it will indicate which pathways are implicated in drug response.

Currently, studies of biomarkers have been mostly performed in the field of candidate genes, and it has to be acknowledged that the overall yield has been disappointing. The few robust results have been in relation to side-effects, e.g. association of the DRD3 variant with extra-pyramidal symptoms, but no single biomarker can be considered to be strongly associated with drug efficacy. This lack of success is not surprising as most studies focused on only a few candidate genes, mainly dopaminergic receptors and transporters and serotoninergic receptors, not taking in consideration the large number of potential pharmacodynamic and pharmacokinetic targets which may be implicated in antipsychotic action either in terms of efficacy or tolerance. The second main weakness of these studies has been the quality of the clinical assessment. Drug response has been mostly measured retrospectively in very inhomogeneous cohorts mixing together first psychotic episodes and chronic schizophrenia patients. Finally, studies to date have examined the usual candidate genes, but have ignored the vast amount of literature and determinants known to influence pharmacokinetic and pharmacodynamic aspects. To overcome this we will use a special pharmaco-genetic (PG) chip developed at INSERM. This PG-chip, which is dedicated to genotyping 15,000 polymorphisms, makes it possible to reconstruct the haplotype diversity of 1292 genes implicated in the metabolism, transportation, and targeting of drugs as well as to genotype the functional polymorphisms that are indispensable for certain genes which are not included in GWA chips. This PG chip includes genes implicated in drug metabolism, drug transport, inflammation (including 158 HLA genes), metabolism, apoptosis, inflammation, chemokines, cytokines, brain receptors and other proteins, signal transduction (carcinogenesis), and DNA repair. The OPTIMISE trial will make it possible to overcome the limitations of the previous
studies.

Recent genetic data have suggested a preponderant role of innate and adaptive immune system in the vulnerability to schizophrenia (Schizophrenia Working Group of the Psychiatric Genomics, 2014; Garcia Bueno et al., 2016; Leboyer et al., 2016) or in antipsychotic treatment response (Le Clerc et al., 2015). Moreover, an increase of inflammatory biomarkers during psychotic episodes has been widely reported in patients with schizophrenia, mainly during acute episodes, suggesting these baseline immune-inflammatory disturbances may be biomarkers of interest for future anti-inflammatory add-on therapy (Fond et al., 2014). Studying inflammatory biomarkers in untreated first episode psychosis is relevant, as antipsychotic drugs are known to influence cytokine levels (Fond et al., 2015). S100B is also of interest, being a calcium-binding protein that is mainly produced by glial cells. S100B levels in serum and plasma are significantly increased in schizophrenia, both in untreated and treated patients (Aleksovska et al., 2014). It has been suggested that S100B may be involved in in inflammatory responses, and that it may be a marker for disrupted blood-brain barrier integrity. The extent to which S100B levels are related to the response to treatment in schizophrenia is unknown.

The detection of circulating autoantibodies against the glutamatergic NMDA receptor (NMDAR-Ab) has been highly debated in psychotic patients. Although there is no clear explanation, the difference between the methods used to detect the autoantibodies is widely accepted as a factor for heterogeneous outcomes. Proteomics in relation to treatment response (pharmacoproteomics) is an emerging field. Holmes et al. (2006) found that metabolic profiles showed a highly significant separation of patients with first-onset schizophrenia from healthy controls. That short term treatment with antipsychotic medication resulted in a normalization of the disease signature in over half the patients indicates the potential for biomarkers as monitors for clinical treatment. However, the Holmes et al. (2006) sample was small, the treatment not fully controlled and the drugs used had multiple modes of action. Metabolomics can also provide valuable information about disease pathogenesis and result in metabolic signatures that could be developed as biomarkers for disease and progression. Pharmacometabolomics is emerging as a new field that could complement pharmacogenomics by providing precise intermediate phenotypes for drug response. Metabolomics could add significantly to our understanding of both pharmacokinetic and pharmacodynamic properties of drugs.

The final promising biological predictor investigated in the current trial is regional brain glutamate levels, as measured using MR Spectroscopy. A proportion of patients with schizophrenia do not respond well to treatment with antipsychotic drugs. Previous research shows that the level of glutamate in the brain differs between patients who have or have not responded to antipsychotic treatment. This suggests that measurement of brain glutamate may be able to predict how well a patient is likely to respond to treatment. However, the previous studies have all been cross-sectional, comparing glutamate in treatment-responders and non-responders. Therefore it is unknown whether these differences in glutamate are present before antipsychotic treatment and can predict response, or, alternatively, whether glutamate changes over time during antipsychotic treatment. If glutamate can predict response to antipsychotic treatment then this would have clinical utility in informing treatment plans.

The study sample
In total, 479 patients were enrolled in the OPTiMiSE study. All the blood samples have been shipped from the different sites to the biobank in Paris. For 403 patients, at least one blood sample is currently available. Among the 403 patients who had blood samples collected on visit 2, 56 patients dropped out before the end of Phase I, 273 patients completed Phase I, 45 completed Phase II, 17 patients completed Phase III. 100% of the available blood samples are now stored at the biobank; a total of 35,612 samples for 789 visits (403 visit 2, 314 visit 5, 55 visit 8 and 17 visit 20). In total, 15,847 serums, 9,448 plasmas, 1,489 buffy coats, 7,339 proteomic plasmas, 1,488 PAXgene tubes for RNA isolation were stored at the biobank. Table 2 summarizes the number of bloods and the number of resulting biological samples that have been collected at each visit among the 479 patients enrolled in the study.

For MRS assessments, the recruitment targets were 63 glutamate scans at baseline and 34 glutamate scans after 4 weeks amisulpride. Glutamate was measured in the anterior cingulate cortex and left thalamus using proton magnetic resonance spectroscopy (MRS). Final recruitment numbers were 72 glutamate scans at baseline and 48 scans after 4 weeks amisulpride. Therefore the recruitment targets were exceeded.

Objective # 1: To test whether glutamatergic markers predict response to first and second line treatments
Our analyses show that there were no significant differences at baseline between patients who were remitters versus non-remitters after 4 weeks of amisulpride treatment, regarding the glutamate levels neither in the anterior cingulate cortex nor in the left thalamus. However, it was found that, on first presentation to clinical services, patients who have higher glutamate have more severe symptoms and worse overall functioning, both at presentation and after 4 weeks of amisulpride treatment. While glutamate levels on presentation do not predict whether a patient will reach remission criteria after 4 weeks amisulpride, glutamate levels in the thalamus may predict the extent to which symptoms may be expected to decline.

Objective # 2: To test if a combination of pharmacogenetic, proteomics- and metabolomic markers can provide clinical valuable predictive value

Transcriptome analyses
Through a pilot study, we identified 85 genes that were differentially expressed (nominal p-value < 0.05) between remitters and non-remitters at inclusion visit and 976 genes were differentially expressed before and after treatment in remitters. Among them, 26 genes were both differentially expressed between remitters and non-remitters and varied after 4-week treatment (Figure 7).

An unsupervised classification analysis using these genes showed perfect discrimination between remitters and non-remitters. Interestingly, among genes identified some have already been associated with antipsychotic response (OLIG1, fold-change=1.72 p=0.01) or have been shown to be differentially expressed in brains of schizophrenic patients (FGFR1, fold-change=0.65 p=0.03). Moreover, several of them are also known to play a role in inflammation (FGFR1 and PLD4, fold-change=1.7 p=0.02) or have been shown to be expressed in microglia, monocytes and macrophages (GPR34, fold-change=1.5 p=0.04). All these consistent results are very promising and suggest that our approach on the whole cohort will allow the identification of a signature of treatment response in patients with first episode schizophrenia and will also reveal molecular mechanisms involved in treatment response.

Inflammatory markers
In a preliminary study, we have investigated whether serum cytokine levels, measured before amisulpride treatment, could be used as biomarkers to predict whether or not First-Episode Psychosis (FEP) patients would meet remission criteria after treatment with this drug. To this aim, we have measured serum cytokine levels in a sub-cohort of 145 men, among whom 95 and 50 were remitters and non-remitters, respectively. Out of the 33 cytokines that we have analysed, 23 (IL-6, IL-7, IL-8, IL-10, IL-12/IL-23 p40, IL-15, IL-16, IL-17A, IFN-γ, TNF-α, TNF-β, CCL2, CCL4, CCL11, CCL13, CCL17, CCL22, CXCL10, SAA, CRP, VCAM-1, ICAM-1, VEGF-A) were above the lower limit of detection in all samples, and 17 (IL-7, IL-12/IL-23 p40, IL-15, IL-16, IL-17A, TNF-α, CCL2, CCL4, CCL11, CCL13, CCL17, CCL22, CXCL10, CRP, VCAM-1, ICAM-1, VEGF-A) exhibited a Gaussian distribution in both remitters and non-remitters as demonstrated using the Kolmogorov-Smirnov test. We then used both standard features (difference of the mean level of individual cytokine between the two classes) and quadratic features (difference of correlation between pairs of individual cytokines between the two classes) to generate a classifier allowing for the discrimination between remitters and non-remitters. To this aim, we used advanced machine-learning (ML) methods based on proximal methods to minimize loss of function. Experiments were performed using training and test sets of 108 and 36 patients respectively, and an iteration of 80 fold. Under these conditions, we have obtained Receiver Operating Characteristic (ROC) curves with an Area Under Curve (AUC) of 97% and 79% for the training and the test set respectively. This promising result demonstrates that serum cytokines could be relevant biomarkers for predicting whether or not FEP patients will meet remission criteria when treated with amisulpride.

S100B levels in samples of serum from a subsample of patients from the OPTIMISE cohort have been examined at King’s College London. Data from this preliminary analysis suggest that S100B levels are correlated with the severity of PANSS positive symptoms at baseline, but that there is no simple relationship between S100B levels and antipsychotic response. However, this may become evident with the application of regression and multifactorial analyses, and with use of the entire patient sample. The final results will be provided to the EC in due time.

NMDAR antibody detection
In order to detect NMDAR-Ab in the sera of first episode psychotic patients of the Optimise cohort, we decided to perform a comparative study using different cell-based assays (e.g. fixed and live HEK cells) in different laboratories (Lyon, France; Kings College, UK; Oxford, UK). In addition, we developed and implemented a new molecular approach, based on the single molecule imaging, to ascertain the presence of circulating NMDAR-Ab once they have been detected by cell-based assays (Bordeaux, France). Using this unique and novel combination of approaches, we could demonstrate that less than 5% of first episode psychotic patients have circulating NMDAR-Ab. We also provide direct evidence that NMDAR-Ab have the capacity to rapidly disorganize synaptic NMDAR, opening the possibility that NMDA-Ab have a direct pathogenic effect on the NMDAR signalling in these patients. The samples which were tested positive have been re-tested to validate these findings. In case samples were tested positive again, the treating physicians of the applicable patients are contacted to discuss these results as well as treatment options for affected patients, as this requires another therapeutic approach than the standard psychosis treatment.

Metabolomics and proteomics
Due to the logistical complexity of collecting the blood samples from all participating centers at the biobank in Paris and subsequently redistribute part of them to German and UK laboratories, it was not possible to report the preliminary results in the current report. Due to the long recruitment period into Work Package 2, the final samples arrived at the Paris biobank beginning of July 2016. The sample analyses are currently being performed and will be reported to the EC as soon as possible. Although the results of metabolomics analyses are not yet received, a pilot study of the applicable Work Package 5 leader suggest that multiple deviations in plasma of schizophrenia patients may be identified, associated with aberrations in biosynthetic pathways.

Conclusion Work Package 5
Although the analyses related to metabolomics and proteomics have not been completed yet, several biological markers resulted from our preliminary analyses on other biomarkers, including a combination of serum cytokines that allow a prediction of amisulpride response (defined as meeting remission criteria) with an accuracy of 80%. In addition, some very promising biomarkers, such as the expression level of OLIG1, FGFR1, SYT2, CTSZ, PLD4 or GPR34, came out from transcriptome analyses and could be used both to predict remission and to be monitored in patient’s blood during treatment. Some of these biomarkers suggest a role of inflammation in the mechanism of response in patients with first episode schizophrenia. Glutamate levels on presentation did no not have predictive value for remission status after 4 weeks amisulpride, however, glutamate levels in the thalamus may predict the extent to which symptoms may be expected to decline. Once the proteomics and metabolomics analyses have been completed, machine-learning techniques can be applied to the complete set of biological markers to determine which combination of markers provides the best prediction for remission, with potential clinical utility. These results will be submitted to the EC as soon as they become available.

Potential Impact:
Part of the envisaged success of the OPTiMiSE project is the effective facilitation of innovation, exploitation and dissemination of findings from the project. OPTiMiSE regards dissemination and exploitation as a linked continuum. They both have a common generic goal: to successfully translate newly gained knowledge to practical application end-points. The key goal of the dissemination activities is to achieve a direct, significant and durable impact on schizophrenia treatment in Europe and beyond.

The results obtained in component A of the project (optimizing current treatment) will provide us with clear, directly applicable, and evidence-based guidelines that will enable us to:
1. Increase treatment response: OPTiMiSE is the first randomised double-blind trial comparing a mechanistically based stay-or-switch algorithm as the second step in the treatment of schizophrenia and the first systematic, large-scale, study assessing the use of clozapine in non-responding patients within the first 10 weeks of their treatment initiation. OPTiMiSE has provided clear indications for treatment guidelines; patients suffering from a first episode of psychosis have a good chance for meeting relatively strict remission criteria when amisulpride treatment is initiated. This is in line with previous scientific findings, which should be translated into local, national and international treatment guidelines. In addition, OPTiMiSE has demonstrated that there is no clear advantage for patients who do not respond sufficiently after 4 weeks of amisulpride in either switching them to an antipsychotic with a different receptor-binding profile versus continued treatment on amisulpride for another 6 weeks. Various current guidelines advise to switch to an antipsychotic with a different receptor binding profile when a patient does not respond sufficiently after a few weeks of treatment, even though this has not been supported by scientific data. OPTiMiSE findings do not support such an advice. Finally, even though remission rates are not very high for first episode patients who initiate clozapine after 10 weeks of treatment, a substantial reduction in psychotic symptoms is demonstrated. This finding suggests that initiating clozapine in a relatively early stage of the illness may benefit non-remitters substantially.
2. Increase treatment adherence: our study on psycho-educational interventions and the use of modern information and communication technologies has led to a novel, directly applicable, tested psychoeducational package to improve adherence and prevent relapse in first-episode schizophrenia patients. Using such a package could further improve patients who are thought to have improved enough already, by responding well to pharmacotherapeutic treatment. The Consortium is awaiting the final results; in case of positive findings, this compliance package can be easily implemented in normal daily practice.

The combination of both approaches can lead to optimisation of current treatments in schizophrenia, with increased treatment effectiveness possibly leading to a higher cost-effectiveness. Once all results are final, health-economics analyses are considered in order to provide insight into the level of (increased) cost-effectiveness.

A third relevant aspect for treatment guidelines, included in component A of the project (optimizing current treatment), was to prevent unnecessary treatment by ruling out organic causes in first-episode schizophrenia patients presenting with symptoms of schizophrenia, therefore preventing unnecessary costly antipsychotic treatments. Our findings indicate a very low prevalence of organic causes for psychotic symptoms. Therefore, implementation of neuroimaging at first admission of first episode patients, in order to rule out organic causes, does not seem to be supported by evidence gathered in the current project, although opinions are still divided. A future health-economics analyses is expected to provide more insight into cost-effectiveness of implementing neuroimaging at first admission.

Second, in component B of OPTiMiSE (exploring novel therapeutic options) our aim was to lay the foundation for personalised medicine in schizophrenia treatment. The expectation was that knowledge gathered in OPTiMiSE on the predictive power of blood-based molecular biomarkers, neurochemicals and neuro-imaging markers could lead to new avenues for individualized and reliable prediction of treatment outcome while our novel insight into the underlying mechanisms of schizophrenia, will lead to the development of new antipsychotics in collaboration with industrial partners. Although the development of a new antipsychotic (cannabidiol, WP4) was not successful, a strong foundation is built for continued research on the efficacy of this potential candidate.

The blood sample analyses conducted so far have provided promising results; we have high expectations regarding the remaining analyses which are currently conducted (i.e. proteomics and metabolomics). Once all results are final, our findings on blood based markers will be combined with our findings in neuro-imaging (both MRI and MRS) in order to develop a prediction model, which can even be extended with demographic data and clinical assessments to increase predictive value. Such a tool can be utilized to optimise personalized medicine in normal daily practice. Depending on the level of predictive value that can be achieved, the existence of this model will be widely disseminated towards all applicable stakeholders, such as the scientific community (e.g. European College of Neuropsychopharmacology, National Schizophrenia and Mental Health Networks in European countries, Schizophrenia International Research Society, Association of European Psychiatry), health care organisations, health care policy makers, treatment guideline committees as well as national and international patient and family associations (e.g. European Network of Users and Survivors in Psychiatry, National Patient Organisations in EU countries, EUFAMI). In addition, the blood markers results are expected to provide new insights into aetiology which, depending on the results, could lead to new directions for psychopharmacological treatment, in collaboration with specific stakeholders (e.g. pharmaceutical industry, regulatory agencies).

List of Websites:
The project website is located at www.optimisetrial.eu and provides information on the background, purpose and setup of the project, as well as a status update on the study progress and Beneficiaries involved in the project, to the general public including patients, clinicians and researchers. The deliverables of the project are explained and the Work Package leaders are introduced. Project-related events are listed and a newsletter is available, to which people can subscribe.

The study website also has a password-protected section, which includes the link to the electronic Case Report Form (eCRF) and enables the Consortium to share any documents of interest to the study: for instance, the protocol, Standard Operating Procedures and Manuals, documents supporting the practical conduct of the clinical trial, questionnaires, multiple training videos, Annual Safety Reports to be submitted to relevant authorities, blood draw forms, Frequently Asked Questions, meeting minutes, slides to promote the study locally and Serious Adverse Event reports. This section of the website facilitates an easy and fast collaboration between Beneficiaries in the conduct of the clinical trial, and contributes to the high quality in data collection.

CONTACT DETAILS:
Sponsor / Coordinator
René S Kahn
[e-mail]
University Medical Center Utrecht, Department of Psychiatry
Heidelberglaan 100, Utrecht, The Netherlands
Phone: [phone]
Fax: [phone]

AUSTRIA
Wolfgang Fleischhacker
[e-mail]
Department of Biological Psychiatry, Innsbruck University Clinics
Anichstrasse 35 A-6020 Innsbruck, Austria
Phone: [phone]
Fax: [phone]

BULGARIA
Luchezar Hranov
[e-mail]
University Hospital of Neurology and Psychiatry ‘St. Naum’
1, Louben Roussev str., Sofia 1113, Bulgaria
Phone: [phone]
Fax: [phone]

CZECH REPUBLIC
Jan Libiger
[e-mail]
Psychiatrická klinika LF UK, Fakultní nemocnice
CZ – 500 05, Hradec Králové, Czech Republic
Phone: [phone] or [phone]
Fax: [phone]

DENMARK
Birte Glenthøj
[e-mail]
Research Center for Neuropsychiatric Research
Ndr. Ringvej, DK-2600 Glostrup, Denmark
Phone: [phone]
Fax: [phone]

FRANCE
Marion Leboyer
[e-mail]
Affiliation # 1: Institut National de la Santé et de la Reserche Médicale (INSERM)
Délégation Régionale Paris 12, Hôpital Henri Mondor
51, avenue du Maréchal de Lattre-de-Tassigny, 94010 Créteil Cedex, France
Phone: [phone]
Fax: [phone]

Affiliation # 2: Fondation FondaMental
Hôpital Albert Chenevier – Pôle de psychiatrie
40 rue de Mesly 94010 Créteil cedex, France

Affiliation # 3: Université Paris-Est Créteil
61 Avenue du Général de Gaulle
94000 Créteil, France

GERMANY
Andreas Meyer-Lindenberg
[e-mail]
Central Institute of Mental Health
J 5, D-68159 Mannheim
Phone: [phone]
Fax: [phone]

Stefan Leucht
[e-mail]
Technische Universität München (TUM)
Ismaningerstrasse 22, 81675 München
Phone: [phone]

Dan Rujescu
[e-mail]
Martin-Luther-University of Halle-Wittenberg, Medical Faculty
Universitaetsplatz 10,
06099, Halle

Thomas Illig
[e-mail]
The Helmholtz Center Munich
National Research Center for Environmental Health (HMGU)
Ingolstaedter Landstrasse 1, D-85764 Neuherberg
Phone: [phone]
Fax: [phone]

Berend Malchow
[e-mail]
Ludwig-Maximilians University München
Nussbaumstrasse 7, 80336 München
Phone: [phone]
Fax: [phone]

Sabine Herpertz
Ruprecht-Karls-University Heidelberg
Medical Faculty Heidelberg
Im Neuenheimer Feld 672, 69120, Heidelberg

ISRAEL
Michael Davidson
[e-mail]
Sheba Medical Centre Department of Psychiatry
Tel Hashomer, 52621
Phone: [phone]
Fax: [phone]

Mark Weiser
[e-mail]
Sheba Medical Centre Department of Psychiatry
Tel Hashomer, 52621
Phone: [phone]
Fax: [phone]

ITALY
Silvana Galderisi
[e-mail]
University of Naples SUN, Department of Psychiatry
Largo Madonna delle Grazie 1, 80138 Naples, Italy
Phone: [phone]
Fax: [phone]

POLAND
Janusz Rybakowski
[e-mail]
Department of Adult Psychiatry, University of Medical Sciences
Szpitalna 27/33, 60-572 Poznan, Poland
Phone: [phone]
Fax: [phone]

ROMANIA
Ilan Gonen
[e-mail]
Tangent Data
Calei Victoriei 81 -93, Scara C ap. 57, 012076 Bucharest, Sector 1
Phone: [phone]
Fax: [phone]

SPAIN
Celso Arango
[e-mail]
Servicio Madrileño de Salud (SERMAS
Dr. Esquerdo 46, 28007 Madrid, Spain
Phone: [phone]
Fax: [phone]

SWITZERLAND
Gregor Berger
[e-mail]
Clienia Schlössli AG, Privatklinik für Psychiatrie und Psychotherapie
Schlösslistrasse 8, CH-8618 Oetwil am See/Zürich, Switzerland
Phone: [phone]
Fax: [phone]

UNITED KINGDOM
Shitij Kapur
[e-mail]
King’s College London, Dept. of Psychological Medicine
PO Box 53, De Crespigny Park, Denmark Hill
London SE5 8AF
Phone: [phone]
Fax: [phone]

Philip McGuire
[e-mail]
King’s College London, Dept. of Psychological Medicine
PO Box 67, De Crespigny Park, Denmark Hill
London SE5 8AF
Phone [phone]
Fax [phone]

Shôn Lewis
[e-mail]
University of Manchester
3rd Floor East University Place, Oxford Road
Manchester M13 9PL
Phone: [phone]
Fax: [phone]

final1-figures-and-tables-for-main-s-t-results.pdf

Related documents

• final1-figures-and-tables-for-main-s-t-results.pdf