ABSTRACT
BACKGROUND: The Western Australian Family Study of Schizophrenia (WAFSS) has conducted genetic epidemiology studies of schizophrenia in Australia for two decades. Recently the WAFSS practices were adopted at the National Centre for for Mental Health in Mongolia, with a view tocollecting comparable data. Like the cited projects (supra), we are cognizant of the dangers of multi site data collection. We replicate common practices, such as training manuals and common site training and refreshment (CCRN WHO training centre). However in international (possibly multilingual) collection and pooling, identical assessment is difficult, it is impossible to replicate endophenotype instructions verbatim (Calkins 2007), and identical recording equipment may not be available indisparate sites. At the very least the data must be compared separately, with the option of weighting,before the pooling for genetic analysis. The use of endophenotypes (Gottesman& Gould) is well established in schizophrenia research for genetic analysis () as well as in more general neuroscience biomarker approaches. The use of electrophysiological markers, and particularly Event-Related Potentials (ERPs) is a well developedaspect of this approach (BraffDL, 2007, TuretskyBI, 2009). Electrophysiological endophenotypes include (inter alia) the Mismatch Negativity (MMN), P50 suppression ratio (P50), auditory oddball P300 (P300), and Antisaccade (AS) tasks. In this study, we seek to follow the multi centre quality assurance examples for pooled data on a smallerscale. This report details the validation of compatibility between the Western Australian Family Study of Schizophrenia (WAFSS) dataset (Perth, Australia), and a pilot dataset from the National Centre for Mental Health (NCMH) in Ulaanbaatar, Mongolia. The working hypothesis is that the psychiatric and endophenotype profiles in the two datasets are sufficiently similar to allow data ompatibility for genetic analysis.METHODS: The Mongolian version of the DIP was developed as part of a joint genetic investigation of schizophrenia between the Centre for Clinical Research in europsychiatry (CCRN) in Perth Western Australia, and the National Center of Mental Health (NCMH) in Ulaanbaatar, Mongolia.The DIP is a semi-structured interview for psychosis for use in epidemiological and clinical settings (CastleD, 2006). It is designed to provide a diagnosis, as well as to assess symptom profiles (present state, past year and lifetime), social functioning, disablement, and service utilisation. It was developed specifically for the National Mental Health Survey – Low Prevalence (Psychotic) Disorders Study(Jablensky et al, 1999, 2000), and has been translated to Italian (RossiA, 2010), Norwegion (SkorvenCS, 2010), and to Mongolian in 2012. The process started with the translation of the original English language version (Castle et al., 2006) by an experienced bilingual psychiatrist (GE) from the NCMH whose native language was Mongolian. Layout and formatting of the document were preserved. It was then back-translated by a non medical,tertiary educated professional, whose native language is Mongolian, but is now resident in Perth. The back-translation was reviewed by an original author (AJ) and experienced practitioners (GP). Grammatical and syntactical discrepancies were resolved directly with the original translator. Event Related Potentials To replicate the WAFSS ERP approach at NCMH, a new portable ERP recording system was deployed. This decision was based on several considerations: a) the WAFSS system could not be taken out of service; b) an identical system could not be replicated due to the age of the components; c) an equivalent system would be too substantial for easy, cost effective transport; d) the system was expected to be used in multiple sites in Mongolia; e) the same system was expected to be used in other Australian projects.The Portable ERP system uses NuAmps, with a hardware selected reference at the FPz location. While the ear references A1 and A2 were recorded, the mathematically re-referenced data is not the same as directly linking ears. (Citation ****). Instead the data was analysed as recorded, with cognizance traces (instead of 20) could not be used. This marks a variation from the original WAFSS processing. Instead of artifact rejection on any trace, only the relevant trace (Fz, Cz, Pz) was used for each ERP (MMN, P50, P300). Endophenotypes The ERP endophenotypes are clearly continous variables, and analysed with general linear modelling. Two tailed significance testing was used for between cohort comparisons, since there is no a priori indication which cohort would have the higher values. Single tailed testing was used in comparing Proband (Pb) and Control (Ctl) groups within the same cohort, as thedirection of any difference is well established.RESULTS: DIP The structure of the diagnostic module (DIP-DM) follows the Operational Criteria for Psychosis, OPCRIT, version 3.31 (McGuffin et al., 1991; Williams et al., 1996) 90-item checklist. It can be used to generate diagnoses according to the criteria of ICD-10 (World Health Organization, 1993); DSM-IV (American Psychiatric Association, 1994); the Research Diagnostic Criteria (Spitzer et al., 1978), and others. The summary of diagnoses (ICD-10 and DSM-IV) generated for each cohort are shown in Figure 1. Diagnostic distribution (%) of 30 interviewed cases from NCMH and 201 cases from the WAFSS cohorts, according to the DIP diagnostic algorithm, by diagnostic classification system. To facilitate omparisons between different criteria systems, Castle (2006) escribes aggregated diagnostic classification descriptors (with reservations) that are used in Figure 1. Greater detail of the DIP responses that support these descriptors is shown for similiarly aggregated questions in Figure 2. aMicrovolts for MMNAmp and P300Amp, numeric forothers.bFor MMN, P50, and AS, but not P300, the raw mean (notabsolute value) for the Pb and Fm groups are higher thanthat of the Ctl group. cEqual variances not assumed.Endophenotype values were each significantly “worse” inthe proband group of the NCMH cohort, for MMN (t=1.65;p=0.05), P300 (t=-2.02; p=0.02) and AS (t=2.12; p=0.02).The comparable values from the WAFSS cohort showed thesame behaviour for MMN (t=4.52; p<0.01), P300 (t=-3.35;p<0.01) and AS (t=3.93; p<0.01). The P50 endophenotypedid not show a significant difference between clinical groups in either NCMH (t=0.20) or WAFSS (t=1.12) cohort. DISCUSSION: This comparison has shown that there is not a significant difference (α= 0.05) between the NCMH and WAFSSpopulations (patient and control). This outcome is deemed sufficient to allow pooled analysis of genetic and electrophysiological data in future studies. It is acknowledged that the outcome does not show that the two populations are the same. Questions of international comparison (McGrathJJ, 2006) in incidence and prevalence, of mental illness and particularly of schizophrenia are eschewed. These were not the purpose of the study. Our experience from this study, as distinct from analysis, is that situational variation in equipment, protocol and recruitment likely outweigh any cultural differencesin epidemiology. The absolute value of the lectrophysiologicalendophenotypes was different between the two sites, butthe relative values were the same. The control group showed“better” responses than the patient group, with similareffect size. Moreover, the patient clinical profile was also slightly different. The incidence of neuroleptic medication was a substantial uncontrolled factor. The question becomes how to deal with these differences.In combining population groups, the data can be discarded,equalized, or transformed. Describe each. We seek to standardize comparisons between populations by transforming data by scaling prior to genetic analysis.Absolute value The raw amplitude data for both ERP eatures (MMN, P300) is significantly lower from the Mongolian cohort in both Patient and Control groups. Endophenotype characteristics.ScalingWhile the difference in absolute values precludes directlycombining data from different cohorts, the consistentendophenotype characteristics allows one possiblemethod to further genetic investigation of continuousendophenotype variables. The results are expected toderive from a combination of technical, situational, clinicaland endophenotype factors. Each of these factors could befurther investigated individually. However, if a combinedendophenotype analysis is even theoretically acceptable,then the endophenotypebehaviour in different cohorts hasto be defined as identical, and the standardized measuresfrom equivalent Control groups must be equal. If the WAFSScontrol group is considered as the standard in this study, then the scaling factors for the NCMH cohort are 13.5 (MMN), 1.0 (P50), 2.5 (P300) and 0.6 (AS).SUMMARY: The consistency in endophenotypebehaviour betweencohorts legitimizes the application of the genetic approachin Mongolia. DNA extraction and analysis for this cohort iscontinuing and, although for smaller numbers, preliminaryresults can be compared with the Australian cohort.