Open Collections

UBC Theses and Dissertations

UBC Theses Logo

UBC Theses and Dissertations

A vulnerability-stress model for the course of schizophrenia ? Erickson, David Harry 1994

Your browser doesn't seem to have a PDF viewer, please download the PDF to view this item.

Item Metadata


831-ubc_1994-893524.pdf [ 4.05MB ]
JSON: 831-1.0099205.json
JSON-LD: 831-1.0099205-ld.json
RDF/XML (Pretty): 831-1.0099205-rdf.xml
RDF/JSON: 831-1.0099205-rdf.json
Turtle: 831-1.0099205-turtle.txt
N-Triples: 831-1.0099205-rdf-ntriples.txt
Original Record: 831-1.0099205-source.json
Full Text

Full Text

A VULNERABILITY-STRESS MODEL FOR THE COURSE OF SCHIZOPHRENIA 7byDAVID HARRY ERICKSONB.A. (Honours), University of Regina, 1978M.A., Simon Fraser University, 1984A THESIS SUBMITTED IN PARTIAL FULFILLMENT OFTHE REQUIREMENTS FOR THE DEGREE OFDOCTOR OF PHILOSOPHYinTHE FACULTY OF GRADUATE STUDIESDepartment of PsychologyWe accept this thesis as conformingo the required standardNovember 1993David H. Erickson7 95In presenting this thesis in partial fulfilment of the requirements for an advanceddegree at the University of British Columbia, agree that the Library shall make itfreely available for reference and study. I further agree that permission for extensivecopying of this thesis for scholarly purposes may be granted by the head of mydepartment or by his or her representatives, It is understood that copying orpublication of this thesis for financial gain shall not be allowed without my writtenpermission.(Signature)Department of_______________The University of British ColumbiaVancouver, CanadaDate_______DE-6 (2/88)1].ABSTRACTDespite a prevailing paradigm that emphasizes an interactionof vulnerability and stress to account for the etiology ofschizophrenia, diathesis—stress models of subsequent course andoutcome of this disorder are rare. Even the simpler stress—process model, where the influence of stressors is mediated bysupportive social relationships, has received little attentionin studies of the course of schizophrenia.The objective of this study was to assess the followingcomponents of a diathesis—stress model as they predict the five-year outcome of first-episode schizophrenia: (1) stressful lifeevents; (2) supportive social relationships; (3) brain lateralventricle size; and (4) smooth pursuit eye movements.As part of the Greater Vancouver M.A.P. Project, we recruitedfirst-episode DSM-III schizophrenia and affective psychosispatients. At intake to the study, their social relationships,smooth pursuit eye movement function, and brain ventricle sizewere assessed. Life events in the previous year were measuredat intake; events over the following 18 months were assessed intwo later interviews. Five years later we assessed outcome,using a global rating of social and occupational functioning.Descriptive results showed substantial variability within theschizophrenia group at intake and outcome. The trajectory ofadaptive functioning over time was remarkably similar for theschizophrenic and affective psychosis groups. Of the four11]hypothesized predictors, only social relationships wereassociated (p=.O3) with five—year outcome. The number of lifeevents was not associated with five—year outcome, nor was eitherof the biological risk factors. As a result, the predictorvariables could not be combined in either a stress—process modelor a vulnerability—stress model of the course of schizophrenia.That social relationship variables are associated with five-year outcome supports earlier findings regarding 18-monthoutcome, including the differing predictive roles for family andnonfamily relationships. The absence of hypothesized resultsfor the life events data probably indicates that too much timehad passed between outcome and the events as measured. Finally,that brain ventricle size and eye-movement dysfunction predict18-month but not five—year outcome may indicate that impairmentdue to biological factors is expressed only in the early stagesof schizophrenia.ivTable of Contents(1) Abstract ii(ii) Table of Contents iv(iii) List of Tables vi(iv) List of Figures viiiCv) List of Appendices ix(vi) Acknowledgements xI. INTRODUCTION 1II. OBJECTIVES 4III. BACKGROUND AND LITERATURE REVIEW 6Social Factors in the Course of Schizophrenia 9Stressful Life Events 9Social Relationships 18Stress—Process Models 24Biological Factors: I. Lateral Ventricle Size 27CT Imaging and Ventricle Size: General Issues . . 29Lateral Ventricles in Schizophrenia 31Correlates of Ventricle Size 37Ventricle Size and Affective Disorders 46Summary: VBR as a Predictor of Outcome 50Biological Factors: II. Smooth Pursuit Eye Movements . 53Systems Governing Eye Movements 53Recording and Scoring SPEM 55Theoretical Requirements for a Genetic Marker . 57Correlates of SPEM Dysfunction 60Summary 64General Issues in the Prediction of Course and Outcome 66Sociodemographic and Premorbid Factors 66Duration of Onset and the Course of Schizophrenia . 67Predictors of Short— vs. Longer—Term Outcome . . 71Current Vulnerability—Stress Models 73Summary and Discussion of Literature Review 77VIV. METHODS: THE MAP PROJECT 83Samples 83Procedures 85Measures 91V. HYPOTHESES AND ANALYSES 103Hypotheses 103Analyses 104Formal Statistical Considerations 113VI. RESULTS 114Phase 1. Methodological and Descriptive Results . . 115Phase 2. The Independent Role of Predictor Variables 132Building Models 156Summary of Results 156VII. DISCUSSION 159Methodology 159Descriptive and Preliminary Results 166Predictive Results 172Models of Course and Outcome 190Significance of the Project 191Limitations 192Future Research 195VIII. REFERENCES 199viList of Tables1. Stressful Life Events, Onset and Relapse 132. Social Relationships and Schizophrenia 233. Ventricle Size and Premorbid Adjustment 394. Lateral Ventricle Size and Prognosis 435. Ventricle Size and Affective Disorders 476. Onset Characteristics and Outcome in Schizophrenia . . . 697. Summary of Measures 878. Sample Size at Three Stages of Intake, by Diagnosis . . . 939. Interview Schedule for Social Interaction: Scale Items . 9510. Summary of Proposed Analyses 10511. Use of Rules in Selecting for Analyses Data FromMultiple Sources Reflecting Five-Year Outcome 11712. Sources of Five-Year Follow-up Data Used for Analyses . 11913. Intercorrelations Among Social RelationshipMeasures, for Two Diagnostic Groups 12314. Factor Analyses of Social Relationship Measures:Communality Estimates and Loadings For ‘Quality’ and‘Quantity’ Factors for All Cases Combined and by Diagnosis . 12515. Sample Characteristics: Demographics and BaselineFunctioning by Diagnosis 12916. Proportions of Patients in Three Categories of AdaptiveFunctioning Over Time, by Diagnosis 13317. Correlations Between Five-Year Outcome and DemographicCharacteristics and Baseline Functioning 13418. Social Networks, Social Resources, and Perceived Support atIntake for Two Groups 13719. Correlations Between Five-year Outcome and Six ‘Ratings ofSocial Relationships, for Two Groups and All Cases Combined. 13920. Hierarchical Regression: The Collective Role of SocialRelationships in Predicting Five-year Outcome 141vii21. Hierarchical Regression: The Unique and Independent Roleof Social Relationships in Predicting Five—year Outcome . . 14422. Number of Life Events at Three Points in Time,by Diagnosis 14823. Correlations Between Five-year Outcome and the Logof the Number of Life Events, by Time and Diagnosis . . . . 15024. Eye-Tracking Error and Ventricle-to-Brain Ratio, byDiagnosis 15325. Correlations Between 18-Month Outcome and the Numberof Life Events or Baseline Characteristics, By Diagnosis . . 17626. Two Regression Equations: Stressful Life Events andBaseline Characteristics Predicting 18—Month OutcomeSeparately for Two Diagnostic Groups 178viiiList of Figures1. Nuechterlein’s (1987) Vulnerability—Stress Frameworkfor Possible Factors in the Development of SchizophrenicEpisodes 762. Axis V Ratings at Intake and Five Years, by Diagnosis . . 1313. Smoothed Distribution of Eye—Tracking RMS Error Scoresfor 124 First—Episode Psychosis Patients 1544. Axis V Ratings at Three Points in Time, by Diagnosis . . 1685. Life Events in Two Periods Prior to Onset, by Diagnosis . 184ixList of Appendices1. Summary of DSM—III Axis V . .. 213xACKNOWLEDGEMENTSWhereas the title page indicates a single author, in factmany individuals and groups have made important contributions tothis dissertation and this degree. Dianne E. Chappell has madepersonal and professional contributions that will not cease withconvocation. Dr. Morton Beiser has inspired “ a love oflearning and a love of life”, in providing an apprenticeship inempirical research. The expert and patient guidance of Dr.Dimitri Papageorgis has been invaluable in the presentation ofthe ideas described herein, as well as in the completion of achallenging degree program.The MAP Project has been a team effort. It has been a joyto work with Jon Fleming, Janice Husted, Bill lacono, KathyKeetley, Neil Kyle, Tsung—yi Lin, and more recently Grahm Beanand Jia-hui Zhang. Four others helped with five-year data: ValGeddes, Richard Hsu, Elizabeth Kao and Janice McLauglin.The biggest collective contribution is from more than 300patients and 200 members of their families, and 200 volunteersfrom the community. They gave freely of their time, theirthoughts, and their feelings in order that staff of the MAPProject might make a small contribution in understandingschizophrenia.Finally, thanks to the taxpayers of Canada and the NationalHealth and Research Development Program of Health and WelfareCanada for National Health Ph.D. Fellowship #6610-1816-47.1I. INTRODUCTIONThe prevailing paradigm for the etiology of schizophreniainvokes a diathesis-stress model, wherein the presence of both apredisposing biological vulnerability as well as some kind ofpsychosocial risk factor(s) must come together to account for the‘causation’ of the disorder. Many prominent authors (Crow, 1988;Gottesman & Shields, 1982; Zubin & Spring, 1977) have describedconceptual models consistent with the diathesis-stress paradigm.Typically, however, investigators gather evidence to supportfactors from only the biological or the psychosocial realm. Theresults may be conceptually nested within a discussion of adiathesis-stress paradigm, but empirical investigations ofetiology seldom bridge the chasm separating the biological andsocial traditions.Compared to investigations of etiology, integrated“biopsychosocial” models for course and outcome are even morerare. This paucity is curious: such a model of outcome seems evenmore compelling than one regarding etiology, in light of theweight of evidence regarding the efficacy of both psychosocial andsomatic forms of intervention. However compelling, suchintegrated, empirically—based models may not exist: in theliterature, there are no thorough assessments of the concurrentand interactive influence of biological and social factorsaffecting the course of schizophrenia.2IntroductionThis dissertation attempts to provide just such a model ofcourse and outcome. From the biological domain two vulnerabilityfactors are examined, smooth pursuit eye movement dysfunction andbrain ventricle size. There is good evidence that this eyemovement dysfunction is under partial genetic control, and isprobably specific to schizophrenia. Large brain ventricles, onthe other hand, can be seen as a vulnerability factor that isneither familial nor specific. To date, neither of thesebiological factors has been adequately assessed as a predictor ofthe course of schizophrenia, in a manner that examines the uniqueand independent role of each.From the psychosocial domain, I will examine a stress-processmodel of the course of schizophrenia. In particular, I willassess the separate and combined influences of stressful lifeevents and social support as they affect long—term prognosis. Noassessment of such a model has been reported in the literature.Many authors have called for interactive models, both inschizophrenia in particular (Nuechterlein, 1987; Zubin & Spring,1977) and in the prediction of behaviour in general (Bern & Funder,1978; Endler & Edwards, 1988). In this project, I will attempt toexamine not only the concurrent influence of social and biologicalfactors, but also the interaction between some of these predictorvariables.This dissertation is based on a large prospective study ofthe course of first-episode psychosis known as the MAP Project.Begun in 1981 by Drs. Morton Beiser and William Iacono, the study3Introductionis based on a representative sample of 175 patients, recruitedshortly after they experienced their first lifetime episode ofpsychosis. The participants were followed over the next fiveyears, with the last of the five—year follow-up interviews beingconducted in 1989. Most of the patients are in three diagnosticgroups: schizophrenia, and either unipolar or bipolar affectivedisorders (both of which had psychotic features). Beyondrepresentative sampling, a number of methodologicalcharacteristics provide a solid foundation on which to constructan integrated, biopsychosocial model of the course ofschizophrenia.The structure of this thesis is as follows. First, thespecific aims are presented, with the rationale provided by areview of the literature. Subsequently, the methodology of theMAP Project is presented. Next, five hypotheses and relatedstatistical analyses will be outlined, followed by thepresentation of the results. Finally, the discussion section willconsider the larger significance of the present findings.4II. OBJECTIVESThe principal aim of this project is to propose an empirically-based biopsychosocial model of the course of schizophrenia, usingdata from a prospective study of a large group of first—episodepatients. In particular, I wish to examine the followingpredictors of five—year outcome:1. Stressful life events2. Supportive social relationships3. Brain lateral ventricle size (ventricle—tobrain ratio; VBR)4. Smooth pursuit eye movement (SPEM) dysfunctionThe objective is to evaluate the unique and independent role ofeach set of predictor variables. To accomplish this aim,prediction equations will control for the baseline measure offunctioning. Other control variables, such as age, sex,socioeconomic status (SES), and the duration of onset (e.g.,insidious vs. acute) will also be incorporated into the predictiveequations in order to control for the more obvious “thirdvariable” artifacts. These controls are necessary in order tosupport possible causal interpretations of the prospectivecorrelations.To the degree that each of the predictors is related tooutcome, they can be considered concurrently. Thus, a secondobjective is to consider a stress—process model of the course ofschizophrenia, whereby supportive social relationships may mediatethe stressor—distress relationship.A third objective is to construct a diathesis—stress model ofthe course of illness by considering concurrently life events,Objectives 5social relationships, and the two putative indices of biologicalliability, VBR and SPEM. To the degree that each predicts five-year outcome, one can assess the manner in which theirinteractions affect the course of illness.The results from any empirical investigation are always morerobust if replicated. As will be elaborated below, publicationsfrom the MAP Project have already demonstrated that supportivesocial relationships (Erickson, Beiser, lacono, Fleming, & Lin,1989) and brain lateral ventricle size (Katsanis, lacono, &Beiser, 1991) each predict short—term outcome. If the hypothesesregarding the effect of these and other variables on longer-termoutcome are supported, then the final objective of the replicationof previous results may be fulfilled.6III. BACKGROUND AND LITERATURE REVIEWConceptions of the course and outcome of schizophrenia havechanged dramatically in the past 15 years. The traditional viewhad portrayed schizophrenia as a disorder with an inevitablydebilitating outcome (Ciompi, 1988). Beginning in the late1970s, however, investigations in North America (Harding, Brooks,Ashikiga, Strauss, & Brier, 1987; Prudo & Blum, 1987; Strauss &Carpenter, 1977) and Europe (Bleuler, 1978; Ciompi, 1980) havedemonstrated that the outcome of the disorder is not universallydismal. Rather, under most diagnostic systems, only about one—third of afflicted individuals attain the state of chronic,perpetual impairment that resembles the traditional schizophrenicprototype. At the opposite, good—outcome pole, up to one-thirdof cases will experience only an occasional subsequent psychoticepisode, with little inter—episodic impairment. The remainingthird has been shown to have an intermediate outcome. Thus, formany patients and their families in Western societies, the morerecent view of course and outcome has provided hope.Results from cross-cultural studies have stimulated hypothesesabout psychosocial factors that might be responsible forheterogeneity in outcome. Affected individuals in developing, asopposed to industrialized, countries have a more benign course(Murphy & Raman, 1971; Waxler, 1977; World Health Organization[WHO] 1972, 1979). For example, people with schizophrenia inIndia and Colombia had fewer and more brief relapses than thosein Washington, Moscow, and Prague (WHO, 1979). These two sets ofLiterature Review 7findings--heterogeneity within and between cultures--point tosociocultural influences on course and outcome.At the same time, developments in technology have providedthe tools for a revolution in the neurosciences, and inparticular for the study of the etiology of schizophrenia. Theadvent of computerized tomographic technology allowed the study of ‘deep structures’ within the brain. This kind ofviewing had not been previously available, and provided directevidence of changes in brain morphology in schizophrenia.Parallel developments in psychophysiology, also due in partto computer technology, stimulated the search for genetic markersof schizophrenia. Because the study of family aggregationpatterns had not been as fruitful as behavioural geneticists hadhoped (Gottesman, McGuffin, & Farmer, 1987), other investigatorshave pursued the study of genetic markers. These genetically-determined traits-—while not necessarily directly involved indisease etiology——could be used to ‘triangulate’ the presence ofa vulnerability or predisposition in family members.By and large, the biological and psychosocial traditions haveproceeded in parallel. Generally, investigators interested inpsychosocial aspects of schizophrenia have conducted research onthe course of schizophrenia, while those interested in biologicalvariables pursued etiology. One reason for the “two solitudes”is likely that investigators are simply not trained in bothtraditions. The strength of these disciplinary boundaries canserve only to impede progress in understanding the disorder andLiterature Review 8providing aid to people with an often disabling and usuallyunpredictable condition.For at least two decades, the diathesis—stress paradigm hasprovided the conceptual integration of the two traditions.Usually associated with the work of Joseph Zubin (e.g., Zubin &Spring, 1977), the model assumes that varying degrees ofbiologically—based vulnerability exist in a large number ofpeople. Among some portion of those individuals, an accumulationof environmental stressors occurs, resulting in profounddistress. The form that distress takes is determined by thebiological vulnerability, in this case schizophrenia. Thisdiathesis-stress model has been applied in empiricalinvestigations of the etiology, but not in the study of thecourse of schizophrenia.The literature review that follows will examine in detailsome of the concepts described above. In the psychosocialsphere, for example, there are maior divergences in competingnotions of the key elements of social relationships. As well,there are several discrepant traditions for assessing thepresence of stressors. Unfortunately, little prospectiveevidence has accumulated on the psychosocial components of adiathesis—stress model of schizophrenia.Following the review of the psychosocial sphere, I willexamine the literature pertaining to two biological variablesthat likely represent different etiological factors. To thedegree that lateral ventricle size and smooth-pursuit eyemovement dysfunction reflect liability for the initial onset ofLiterature Review 9schizophrenia, they may also represent a continuum of liabilitywith respect to course and outcome of the disorder.Social Factors in the Course of SchizophreniaOne of the more prominent models in mental health, thestress—process model, brings together two sets of psychosocialfactors to account for the etiology of distress. As will be seenbelow, the interaction of stressful life events and supportivesocial relationships has been repeatedly applied to the study ofmany domains of mental health. Nonetheless, the relevantliterature on schizophrenia is relatively limited.Stressful Life EventsIn the stress—process model, stressful life events (LE5)have been proposed as having an adverse impact, either to promptthe initial occurrence of distress, or to adversely affect thecourse of a disorder. The notion is that the accumulation ofstressors at some point exceeds the individual’s capacity to dealwith the multiple demands, and at that point a person becomesdistressed.Definition and measurement of stressors. Stress has beenused to describe both the cause of a decline in mental health,and the effect itself. In the lay idiom, having “too much stressin one’s life” can refer to the presence of unwanted and negativeevents, or it can refer to the distress that follows theLiterature Review 10occurrence of those events. In this thesis, the term ‘stress’will be avoided. Instead, ‘stressors’ will refer to life eventsor other challenges, and ‘distress’ to one’s subjective reactionto the occurrence of life events.There are several concurrent traditions in the study ofstressors. One of these examines the effects of a single andoften catastrophic event, e.g., a natural disaster or injuryfollowing a motor vehicle accident (Henderson & Brown, 1988). Asecond tradition examines the effects of ongoing strain on mentalhealth, such as being physically disabled (Cf. Turner & McLean,1989). A third examines the effects of ‘micro—events’, oftendescribed as ‘daily hassles’ (e.g., Lazarus & Folkman, 1984).Most frequently, though, investigators have assessed thecumulative effects of more ordinary life events. Here, theeffects of losing one’s job, the birth of a child, a change inresidence or the start of an intimate relationship, are assessedwith respect to mental health. In this tradition, studies haverepeatedly demonstrated that the accumulation of these stressfullife events is associated with decrements in mental health(Finlay—Jones, 1988).The measurement of life events typically occurs with the useof a structured checklist. The original instrument developed byHolmes and Rahe (1967) is relatively brief. It includesquestions that ask about the occurrence of 34 LE5 in either thepast 6 or 12 months. Elaborations of this method often applylinear weights in summing the LEs, which are based on therelative recency of each event, and on the subjective weightingLiterature Review 11of the effect or the normative impact of each event (Thoits,1983).A more exhaustive assessment of LEs has included a lengthyinterview about each event (Brown & Harris, 1978; Brown, Sklair,Harris, & Birley, 1973). Here, the participant tells thedetailed story of the occurrence of the event. An interviewerrecords the entire narrative, and then embarks upon an elaborateset of questions regarding each event, with reference to theextent to which it was related to current mood and was within thecontrol of the respondent. This protocol is used for the purposeof interpreting the effects of stressors both at the cumulativelevel and at the level of the individual life event. The choiceof this more thorough interview versus the more brief checklist(described above) depends on the objectives of the researchinvestigation.In order to begin to argue for a causal role, the temporalpriority of life events and symptoms must be established. Innonchronic diseases, it is usually clear whether life eventsprecede or follow the onset of symptoms. In chronic or relapsingdisorders such as schizophrenia, disaggregating the events thatare putative causes of symptoms from those that are the result ofsymptoms is accomplished by classifying events as independent vs.dependent, respectively. A third category, ‘possiblyindependent’, is reserved for events that are not likely to havebeen the result of pre-existing distress.Life events and the onset of schizophrenia. Three studies(summarized in Table 1) have assessed the role of LEs inLiterature Review 12precipitating the initial onset of schizophrenia. Day, Nielsen,Korten, et al. (1987) reported on a cross—national study offirst—admission patients in six countries. In each site,participants compared LEs in the three weeks prior to the firstepisode with those in previous three week periods. In five ofthe research centres, patients reported an excess of independentevents, i.e. those outside of their control, immediately prior totreatment contact. Thus, it appears that an increased number ofLEs may have played a causal role in the onset, as opposed tobeing a result of early signs of illness.An earlier study by Jacobs and Myers (1976) also examinedLE5 among first-admission schizophrenics compared to thosereported by a matched, normal comparison group. They inquiredabout LE5 in the year prior to entry in the study. They foundthat schizophrenics had significantly more LE5 than did normals,although there was no significant difference when the events wererestricted to those rated as independent. Thus, this studyoffers only equivocal support for a possible causal role for lifeevents in the onset of schizophrenia.A third investigation, failed to find an excess of LEs amongschizophrenics. Gureje and Adewumni (1988) compared their firstepisode group to a matched normal comparison group, and enquiredabout events in the previous six months. There were no groupdifferences in the mean number of LEs, nor did the proportion ofeach group reporting at least one LE differ. The negativefindings were also true for the three—month period prior tointake.Table1.StressfulLifeEvents,OnsetandRelapseAuthor(s)SampleResultsmLifeeventsandonsetDayetal.Schiz.patientsin1.Allsites:excessLEsin3wks.priorto(1987)sixcountries,intake,comparedtoprevious3wk.period.2.Moreindep.LEspriortointakein5/6sites.Jacobs&MeyersSchiz.patientsvs.1.AllLEs:schizophrenics>controls.(1976)matchedcontrols.2.IndependentLEa:nogroupdifferences.Gujere&AdewumniSchiz.patientsvs.1.Nogroupdiff.inmeannumberofLEaIn(1988)matchedcontrols.either3or6monthperiodpriortointake.2.Nodifferencebetweengrps.inproportionwhoreportedatleastoneLE.LifeeventsandrelapseVenturaetal.“Recentonset”schiz.pta.RelapsingpatientshavemoreIndependentLEs(1989)interviewedevery8wks.inmonthpriortorelapse,comparedto:over12monthsa)moreremoteperiodsinthestudy,andb)non-relapsingpta.inthesamemonth.AlKhanietal.Acuteandchronicschiz.1.Womenonly:pta.hadmoreLEathancontrols.(1986)pta.inSaudiArabia2.ClusterIngofLEain3weekspriortorelapseonlyamongmarriedfemalepatients.Leffetal.Unmedlc’dchronicschiz.pta.Relapseprecededbyatleast1LEin56%ofpta.(1973)inhigh-vs.low—EEhomesInlowEEhomes,vs.9%inhighEEhomes.Leffetal.Medic’dchronicschiz.pta,83%ofrelapsingpta.vs.25%ofpta.of(1983)allinhighEEhomesnonrelapsingpts.havethreateningLE.I- (.‘)Literature Review 14Overall, it seems that the key to detecting the hypothesizedrole of life events in prompting the initial onset ofschizophrenia may be to use a within-subjects design. When Dayet al. (1987) used this strategy, five of six centres found anexcess of independent events.Life events and relapse. Four studies have provided directevidence to indicate that LEs play a role in relapse. In aprospective study, Ventura, Nuechterlein, Lukoff, and PedersonHardesty (1989) described a group of recent-onset schizophrenicpatients who had just finished a treatment program. Throughoutthe follow-up period, life events, symptoms and adaptivefunctioning were documented. Their results showed that relapsingpatients had more independent LEs in the month prior to thepsychotic episodes, compared with non—relapsing patients. Thesame result was found when relapsing patients were used as theirown controls: There was an excess of LEs immediately prior tothe episode, compared to more remote periods.In an early study, Leff, Hirsch, Gaind, Rhode and Stevens(1973) studied a mixed group of acute and chronic patients, whowere randomly assigned to long-term drug or placebo treatment.Their results showed that a larger number of patients whorelapsed while on medication reported at least one life event inthe previous five weeks, compared to relapsing patients onplacebo. The authors interpreted this to show that schizophrenicpatients who were on medication were “protected against thestresses implicit in uneventful social intercourse, and wereLiterature Review 15unlikely to relapse unless exposed to some additional stress inthe form of one or another life event” (p. 660).In a later study, the same research team (Leff, Kuipers,Berkowitz, Vaughn, & Sturgeon, 1983) compared relapsing tononrelapsing patients, all of whom were living in high EE homesand were on medication. Their results showed that life eventswere more often present in the three weeks prior to relapse,compared either to previous periods or to nonrelapsing patientsthroughout the study. The authors interpreted the results asshowing that acute stressors (life events) had an additionaleffect beyond that represented by chronic strain (high levels ofexpressed emotion).A fourth study also offers positive, but qualified, results.Al Khani, Bebbington, Watson, and House (1986), in a SaudiArabian study, found a higher frequency of LEs among a mixedgroup of acute and chronic schizophrenics, compared to a matchednormal comparison sample--but only among women. They also founda clustering of LEs in the three weeks prior to the indexepisode, but only among married female patients (compared to thecorresponding normal control subgroup).Life events and course. Host studies investigating LEs andschizophrenia have examined the more-or-less immediateconsequences of situational stressors. That is, the role of LEshas been considered as a precursor either to initial onset or torelapse, thereby taking a short-term view of outcome. This usualview of the role of LEs corresponds to a “triggering” model.That is, LEs precipitate onset or relapse, but they do notLiterature Review 16substantially alter the probability that it will happen--only itstiming. In their model, Brown and Birley (1968) even developedformulae to estimate the degree to which LE5 bring forward intime the occurrence of a psychotic episode.A different model might test the effects of LEs on thecumulative course of illness. If it is true that LEs affect thetiming of a psychotic episode, then a cumulative measure ofcourse may show adverse effects. For example, an episode maybegin sooner because of LES, but the timing of its resolution maybe unaffected—-resulting in longer episodes. Alternatively, tothe degree that LEs “bring forward” the timing of an episode, butdo not affect its length, the duration of symptom—free inter-episodic periods would be shorter. In either case,schizophrenics with a relative abundance of LE5 should show apoorer course. In this way, life events assume the status of ahazard, and not simply a precipitant. This model has not yetbeen empirically tested*.Limitations in the LE area. There are two major issues thatare common to all of the LE studies, but they have yet to beaddressed——either conceptually or empirically. The first issuerelates to the frequent exclusion of slow-onset patients: moststudies (Brown & Birley, 1968; Day et al., 1987; Gureje &Adewumni, 1988; .Jacobs & Myers, 1976; Ventura et al., 1989)* Only recently has there has been an investigation of thecumulative effects of LES over a longer period of time. Hirsch,Bowen, Emami, Cramer, Jolly, and Haw (1993) found that LE5 actedcumulatively to increase the risk of relapse over a 12-monthfollow-up period. To date, their work has been reported only inabstract form.Literature Review 17included only those patients whose onset occurred within 3-6months of referral to the study. If it is true that acute (vs.insidious) onset and the presence of precipitating events clustertogether (Dohrenwend & Egri, 1981), then the practice ofexcluding slow-onset patients may render the apparentlysubstantive findings (outlined above) an artifact of thisexclusion criterion. Only a study that includes the full rangeof insidious and acute onset patients would be able to resolvethis issue.A second significant issue yet to be addressed relates tothe conceptualization and measurement of the onset period. Inthe LE literature, it is now a fundamental requirement thatevents must be dated with respect to the onset of psychoticsymptoms (Dohrenwend & Egri, 1981; Spring, 1981), in order toseparate LEs that may cause onset from those that may be theresult of the disorder. However, one further step is necessary.There remains the possibility that some LE5 are the result ofprodromal symptoms. Such a possibility is particularly cogentwhere the onset of the prodrome is not accompanied by profoundoccupational or social impairment. For example, a working parentmay experience several weeks or even months of social withdrawaland flat affect. Even without florid psychotic symptoms, onewould expect LBs that follow the onset of the prodromal period,such as reprimands in the workplace or arguments with a spouse.Thus, a proper account of onset must include the prodromalperiod.Literature Review 18Finally, studies of life events tend to be based onnaturalistic, longitudinal research designs. Here, assertionsregarding causality rely on prospective correlations ortemporally—based comparisons (e.g., the number of life eventsimmediately before relapse vs. other time periods). However, thedefinitive standards by which causality is demonstrated requireevidence beyond prospective correlations, even if obviousconfounding factors are controlled. Other evidence might bederived from the experimental manipulation of stressors toexamine resultant distress. In general, ethical requirementsprohibit the collection of this kind of evidence.Summary. There is good evidence that an increase in thenumber of life events may play a causal role in the onset of bothinitial and subsequent episodes. In this triggering model, morework is needed, based on research designs that include a broadsample of patients and an adequate assessment of the duration ofonset. Similar conditions are needed to assess an alternatemodel that addresses LEs with respect to course and outcome,wherein the influence of possible mediating variables such associal relationships may be examined.Social RelationshipsA vast literature has accumulated on the effects of socialrelationships in a wide array of physical and psychologicaldisorders. In physical medicine, the presence of supportiveothers leads to increased survival rate after the detection ofbreast cancer and the occurrence of myocardial infarct (Medalie &Literature Review 19Goldbourt, 1976), reductions in the severity of arthritis andchronic pain (Davidson, Bowden, & Tholen, 1979; Porrit, 1979),and decreases in the rate of pregnancy and birth complications(Nuckolls, Cassel, & Kaplan, 1972).In the mental health sphere, greater amounts of supportiverelationships have been shown to have a protective function:Among people experiencing stressful life events, rates ofdepressive and anxiety symptoms are substantially lower for thosewith more social support (Mueller, 1980). While there is animpressive body of evidence for the role of social relationshipsin preventing the occurrence or mitigating the effects of anxietyand mood disorders (Alloway & Bebbington, 1987), there isrelatively little empirical work on the role of social support inschizophrenia. Before reviewing that evidence, however, someclarification of the terminology and concepts is required.The “social support” literature suffers from problems ofdefinition. While most authors share the perspective that thereis something about social relationships that is valuable in termsof resistance to or recovery from distress, agreement stopsthere. There is little consensus about the concept itself, andeven less about the critical elements of social relationshipsthat are thought to be helpful. Most approaches to the study ofsocial relationships fall into three categories: socialnetworks, social resources, and the perception of social support.The social network approach involves the study of thestructural characteristics of a person’s social relationships.The most obvious is the number of people named to the networkLiterature Review 20(Greenblatt, Becerra, & Serafetinides, 1982). The composition ofthe network is also associated with one’s mental health: thepresence of a confidant, often a spouse, is most important (Coyne& DeLongis, 1986). In general, greater numbers of both kin andnon-kin are important to the preservation of good mental health(Schradle & Dougher, 1985), as is the presence of close friends,compared to acquaintances (Henderson, 1988). The networkapproach, then, has shown that a number of structuralcharacteristics are related to good mental health.The transactional approach to social resources focuses only onthose types of relationships that fulfill certain functions.Thus, theorists have tried to categorize the nature or thequality of the transactions that comprise social relationships,and focus on those that are deemed to enhance health status. Forexample, Caplan (1974) emphasizes mutuality and reciprocity ineach of emotional support, cognitive support, and tangibleassistance. Cobb (1976), on the other hand, argues that theessential feature is the provision of informational feedback.Henderson (1984, 1988) offers a more complex taxonomy based onattachment theory (Weiss, 1974).Turner (1983) argues that the transactional approach confoundsthe presence of social relationships with the perception of beingsupported. In his view, social support is always positive: Itis the degree to which one is satisfied with the current socialresources. Since the experience of being supported is alsoaffected by other factors, he argues that controlling forLiterature Review 21constructs such as mastery and locus of control successfullycaptures the perception of social support per Se.There is no consensus in the area on the definition of health-enhancing interpersonal transactions. Empirically, there is noreason to select as best either the network, the transactional,or the perception-based approach. The best approach appears tobe an inclusive one that incorporates the network, transactional,and perception-based elements. For example, it must recognizethe distinction between the support arising from a spouse, goodfriends, and acquaintances. Moreover, an inclusive measureshould assess emotional support, cognitive or informationalsupport, and some index of tangible assistance. If data based onall of these considerations are collected, then supportive socialrelationships have been well—measured.Social relationships and schizophrenia. While the literatureon supportive social relationships and emotional health ingeneral is substantial, that relating to schizophrenia is small.Most authors who have studied the psychotic disorders have statedthat their ultimate interest is in the role of socialrelationships as they influence the course of the disorder, withpossible implications for treatment. In spite of this statedintention, all but two papers have been based on a crosssectional design: unable to comment on the direction ofcausality, most of the published reports can provide no clearimplications for treatment. Nonetheless, they do provide astarting point.Literature Review 22Host investigations in the area (summarized in Table 2) havetaken the network approach, based on cross—sectional researchdesigns. Here, a number of studies (Erickson, Beiser, lacono,Fleming, & Lin, 1989; Garrison, 1978; Pattison, DeFrancisco,Wood, Frazier, & Crowder, 1975; Sokolovsky, Cohen, Berger, &Geiger, 1978; Westermeyer & Pattison, 1981) have reported thatthe networks of schizophrenic persons are smaller than those ofnon-psychiatric comparison groups. This reduced network size isprobably not specific to schizophrenia: in an earlier reportfrom the MAP Project, we reported that a comparison group ofaffective psychotics also had smaller networks, although thisgroup occupied a median position between the schizophrenic andnormal groups (Erickson et al., 1989).Evidence also suggests that the networks of schizophrenicpatients become further reduced over the course of illness(Westermeyer & Pattison, 1981). For example, Sokolovsky et al.(1978) showed that, over a two—year span, network size diminishedby 40%.Schizophrenics also appear to differ in the structuralcharacteristics of their social networks. One of the mostwidely—cited findings is that family members are overrepresented, compared to the networks of both psychiatric andnormal comparison groups (Pattison et al., 1975; Tolsdorf, 1976).We reported that the relative over—representation of kin is notdue to an increase in their absolute magnitude in the network,but rather to the decreased presence of friends and acquaintances(Erickson et al., 1989). In summary, most studies of socialTable2.SocialRelationshipsandSchizophreniaAuthor(s)SampleResults(bPattisonetal.Neuroticvs.psychoticpts.vs.1.Networksize:normals>neurotics>lb(1975)normativecommunitysample.psychotics(mostlyfamily).lbTolsdorf(1976)10first—admissionschiz.pts.1.Size:nosignificantdifference.vs.10medicalcontrols.2.Multiplexity:controls>schiz.3.kininnetwork:Schiz>controls.Garrison(1978)PuertoRicanwomeninNYC:1.Networksofnormativesamplearelargerschiz.pts.vs.controls,innumberandgreaterinmultiplexity.2.Schiz.pts.haveonegoodfriendatmost.SokolovskyChronicschiz.pts.inhotel,vs.Networkcharacteristics:etal.(1978);occupantswithnopsychhistory.a)size---normalcontrols>asymptomaticCohen&schiz.pts.>schiz.withresidualsx’s.Sokolovskyb)multiplexity——samerankorderasabove(1981)butnonsignificanttrend.Outcome:networkmeasurespredictedrehosp’zn,butonlyformildlyimpaired.Westermeyer&Laotianpeasants:schiz.vs.Psychoticvillagershavesmallernetworks,Pattlson(1981)affectivepsychosisvs.&theirsocialrelationshipsarenormalcontrolsmadeupofsmallerandfewer‘clusters’.Ericksonetal.First—episodeschiz.vs.affectiveAtintake:(1989)psychosisvs.normalcontrols.a)Kininnetwork:controls>aff.>schiz.b)Availabilityofsupportfromconfidants:schiz.<affectives,controls.C)Adequacyofsupportfromacquaintances:schiz.<affectives,controls.Predictingoutcome:a)Morenonkinpredictsgoodoutcome(allpts.).b)Forschiz.pts.:moresupportiveacquaint—ancespredictsbetteroutcome,butmorekinwinnetworkpredictspooreroutcome.Literature Review 24relationships and schizophrenia have used a cross—sectionalapproach, where authors found differences between schizophrenicand normal comparison groups.Only two papers have directly assessed the role of socialrelationships on the course of the disorder. One of thosereports (Cohen & Sokolovsky, 1978) concluded that, among chronicpatients, there is evidence that social relationships do have apositive effect on outcome but only among the mildly impaired.Among more severely impaired chronic patients, socialrelationships did not influence outcome. The second paperexamined first-episode patients: Erickson et al. (1989),describing the MAP Project sample, reported that more numeroussocial relationships outside the family appear to have a positiveeffect on course. By contrast, the presence of greater numbersof family predicted a poorer short-term course forschizophrenics. A somewhat different picture was true for theaffective psychotic patients: greater numbers of both kin andnonkin had a positive effect on short-term outcome.Stress-Process ModelsIf research documenting the unique and independent effectsof stressful life events and supportive social relationships onmental health is the first step in building a stress—processmodel, then assessing their simultaneous effects is the nextstep. That second step, building a two—factor stress-processmodel, has been applied repeatedly to the study of many physicaldisorders (Cohen & Syme, 1985). In the mental health sphere, theLiterature Review 25model has been validated in the study of clinical populationswith affective and anxiety disorders, as well as inepidemiological studies of nonclinical populations (Mueller,1980). In each case, two kinds of effects are noted: (1) a maineffect, where the presence of supportive social relationships hasbeen associated with reductions in the number of LE5; and (2) aninteractive or buffering effect, where social relationships areassociated with diminished effects of LEs when they do happen.A three—factor stress—process model has also been applied tothe study of many disorders and conditions. Here, the concept ofcoping is added to the consideration of social relationships asan additional intervening factor in the stressor—distressrelationship (Lazarus & Folkman, 1984).Both the stress-process model and the expanded stress-and-coping model have been the subject of empirical investigation forat least 20 years. Despite hundreds of reports about dozens ofphysical and mental conditions, neither model has ever resultedin a published report about schizophrenia. Neither model hasbeen applied empirically to either the etiology or the course ofschizophrenia. The apparent absence in the literature of thesemodels is indeed curious. It may be that empiricalinvestigations of these models vis-a—vis schizophrenia have cometo nought, and the negative findings have simply not beenpublished.In sum, the best information available on the stress-processmodel as applied to the psychotic disorders Involves separate andparallel results about its components. Regarding the firstLiterature Review 26component, the stressors, investigators of schizophrenia haveused the cumulative index of life events in most published work.There, the objective has been to document the proximalassociation between an increase in the number of stressors andeither initial onset or a subsequent relapse. No report hasassessed the long—term, cumulative effect on course.In the study of other domains of mental health the secondcomponent of the stress—process model, supportive socialrelationships, has frequently been shown to have a preventativeand a buffering role in the face of stressors. In schizophrenia,however, most of the empirical reports of supportive socialrelationships have been limited to findings based on cross-sectional research designs. Here, the results do suggest thatschizophrenic people have smaller social networks, fewer socialresources characterized by supportive transactions, and generallyfeel less supported than other people. In examining the effectof supportive social relationships on the course ofschizophrenia, two longitudinal studies have shown that socialrelationships influence outcome. Even so, differences in samplesand measures mean that replication is still needed.Literature Review 27Biological Factors: I. Lateral Ventricle SizeIn 1976, Eve Johnstone and her colleagues at the Institute ofPsychiatry in London first described the use of a new X-rayimaging technique to study in vivo the structure of the brains ofschizophrenic people (Johnstone, Crow, Frith, Husband, & Kreel,1976). They found that, compared to a medical control group,schizophrenics’ lateral ventricles were moderately enlarged.Since that first study, the cumulative results from more than 100reports using computerized tomography (CT) have provided in vivoevidence that schizophrenia is very likely, in large part, a‘primary brain disease’.The lateral ventricles, also known as the first and secondventricles, are symmetric fluid-filled cavities on each side ofthe brain. Surrounding these cavities are midbrain structuressuch as the thalamus, hypothalamus, hippocampus, fornix, andcorpus callosum. Investigators hoped that any ventricularenlargement detected on a CT film would be a direct clue topathology in surrounding tissue. Previously, other lines ofinvestigation had found evidence of pathology in thesestructures. Never before, however, had a technique beenavailable that would rule out the artifactual explanations of theprevious results by offering in vivo evidence.Most of the published work has been devoted to establishingthe presence of ventricular enlargement, with the aim ofinvestigating the etiology of schizophrenia. As will be seenbelow, many technological and methodological issues have troubledLiterature Review 28the area, but most authorities now believe that ventricularenlargement is present among at least a substantial portion ofschizophrenics, and that it occurs at or before the onset of thefirst psychotic episode. The magnitude and prevalence ofenlargement, as will be seen, are issues that await definitiveresolution. When taken In the aggregate, though, there iswidespread agreement that lateral ventricular enlargement ([NE)is related to etiology.Enlarged ventricles may reflect a relative prominence ofnonfamilial etiological influences in schizophrenia. In supportof the ‘sporadic’ vs. ‘familial’ distinction, four studies havedocumented an association between increased VBR and an absence offamily history of schizophrenia (Pearlson, Garbacz, Moberg, Ahn,& DePaulo, 1985; Reveley, Reveley, Clifford, & Murray, 1982;Sacchetti, Caizaroni, Vita, Terzi, Pollastro, & Cazzullo, 1992;Turner, Toone, & Brett-Jones, 1986). Other studies, however,have failed to find this relationship (Farmer, Jackson, McGuffin,& Storey, 1987; Reddy, Mukherjee, Schnur, Chin, & Degreef, 1990),and one has documented an association in the opposite direction(Owen, Lewis, & Murray, 1989).Because of several limitations, the significance of LyE forthe course of illness is not clear because, for the most part,possibly overlapping relationships between ventricle size andpremorbid functioning, clinical state characteristics, and otherindicators of biological vulnerability have not been controlled.To the degree that these other correlates of ventricle size areLiterature Review 29examined, the significance of increased ventricle size on outcomemay become clearer.CT Imaging and Ventricle Size: General IssuesPrior to the development of CT technology, in vivo imaginghad been done with pneumoencephalography, whereby air wasactually injected into the head via the lumbar subarachnoidspace. X-rays of the head would then allow imaging of theventricular system and cortex. Between 1920 and 1960, manyinvestigations had reported enlargement of ventricles and atrophyof cortical tissue in schizophrenia (Haug, 1962).Since the middle of the 1970s, the technology involved incomputerized tomographic imaging has been the basis of anunprecedented advance in the study of schizophrenia. The newtechnique has offered a noninvasive in vivo view of brainanatomy, free of artifact that had been associated with previousmethods (Seidman, 1983).While CT technology has provided a tremendous impetus toresearch into the etiology of schizophrenia, its major limitationis that CT data index relatively gross anatomy in the brain.This is primarily a function of the limited powers of resolutionin the image. For the past decade there has truly been arevolution occurring in the neurosciences, but the revolution hasnot yet advanced to the stage of mapping in sufficient detail thebrain structures--and their interconnections--that are likelyanomalous in schizophrenia. Advances in technology that allowin vivo study of fibres and tracts are rapidly developing, e.g.,Literature Review 30new structural imaging (magnetic resonance imaging; MRI), as wellas functional mapping (e.g., positron emission tomography [PET],regional cerebral blood flow [rCBF], dynamic MRI)*. Overall, thelimited resolution of the CT image puts a similar limit on theconceptual understanding of the specific mechanism(s) involved.Measurement issues. The changes in ventricle size that arehypothesized to occur in schizophrenia are subtle, and requirespecial attention to measurement and methodology. Unfortunately,many of the earlier studies were based on simple linearmeasurements, where the ventricle was measured at its widestpoint, and compared to the maximum width of the total brain.Methodological studies (e.g., Penn, Belanger, & Yasnoff, 1978)soon showed low correlations between this one—dimensionalapproach and the actual volume of the ventricles. A betteralternative is the measurement of area using a planimeter,whereby the outlines of the ventricles and the brain are traced.The area of the ventricle is then calculated as a percentage ofthe total brain area. This ventricle-to-brain ratio (VBR) issensitive to small differences, and correlates well (r>.95) withthree-dimensional assessments of total volume (Penn et al.,1978).Ventricle size and diagnostic specificity. Enlargement ofthe ventricles can be the result of many causes other thanschizophrenia, and the implications of enlargement are notnecessarily clear. For example, LVE can be permanent, following* The new technologies have superseded the CT approach, but CTbased data are still current with respect to the course ofschizophrenia.Literature Review 31blockage of the flow of cerebrospinal fluid (CSF; TerBrugge &Rao, 1983) or degeneration of brain tissue (Bird, 1982). It canalso be reversible, as is the case with a marked change inelectrolyte balance or poor nutritional status (Benston, Reza,Winter, & Wilson, 1978). Dilation of a ventricle can belocalized, for example following trauma, inflammation, orvascular hemorrhage, or it can be diffuse. The latter is thecase in schizophrenia (Seidman, 1983), but also in Alzheimer’sdisease (Bird, 1982) and alcoholism (Carlen, Wilkinson, Wortzman,& Holgate, 1984).Ventricular enlargement does not necessarily imply tissueatrophy. For example, in Alzheimer’s disease, a change inventricle size is invariably accompanied by tissue atrophy. jp37Contrarily, in Huntington’s disease and hydrocephalus,ventricular enlargement occurs without atrophy (TerBrugge & Rao,1983). A third possibility is that brain tissue can be affectedby some conditions, e.g., cerebral anoxia, without a change inventricle size. In summary, any change in ventricle size is notspecific to schizophrenia, and does not necessarily imply atrophyin brain tissue.Lateral Ventricles in SchizophreniaSeveral measures of brain structure have been studied inschizophrenia. The size of the third ventricle, the corticalsulci, and the cerebellum have been examined in addition to thesize of the lateral ventricles. Only the lateral ventricles haveLiterature Review 32been selected for examination in this study, for the followingreasons:—— due to the limited resolution of the CT image, largestructures are best suited for study—— there is better agreement on measurement issues in lateralventricles, as compared to measurement of thirdventricles, sulci and tissue atrophy-— lateral ventricles are the most frequently studied,and hence best understood-- limited evidence from previous studies suggests a linkbetween LV size and prognosis, unlike other measuresof brain structure-— two other parts of brain structure, third ventricles andmeasures of cortical tissue atrophy, are small,resulting in limited variance, which in turn can beexpected to attenuate any predictive relationshipFor these reasons, lateral ventricle size was the bestavailable index of changes in brain structure, at the time of theinception of the MAP Project.Group differences in ventricle size. Raz and Raz (1990)recently conducted a quantitative review of the more than 100studies that have examined the CT scans of schizophrenic people.Regarding lateral ventricles, the authors found a mean effectsize of 0.70 (SD=.54), based on 53 comparisons between patientsand controls. This effect size corresponds to a 5Th overlapbetween the two groups.In their review, Raz and Raz (1990) classified five of the53 comparisons as outliers, because the effect sizes were largeand discontinuous with those derived from the other 48comparisons. The distribution of 48 studies formed a smooth andnormal curve, suggesting that lateral ventricle size is notLiterature Review 33bimodally distributed for schizophrenic and ‘normal’ people.Removal of the five outliers reduced the effect size to M=.57,which corresponds to a 63 overlap. In Cohen’s (1977)nomenclature, this is roughly midway between a moderate (d=.70)and a small (d=.40) effect size.The method of measurement seemed to make a difference in thefindings. Where two-dimensional planimetry was used, the meaneffect size was significantly larger (M=.76) than with linearmeasures (M=.43). This difference is even more dramatic if theoutliers are removed (.63 vs .23, respectively; Raz & Raz, 1990),although the magnitude of the mean effect size is reduced. Thus,the best estimate from the meta—analysis would appear to be basedon area measures with the ‘outlier’ studies removed fromconsideration: the effect size from such studies would appear tobe approximately 0.60.Timing of ventricular enlargement. Two strategies have beenused to examine the timing of enlargement, with the purpose ofdetermining whether ventricular enlargement is a possible causeor result of the disorder. The first examines ventricle sizeamong schizophrenic people at the time of their first lifetimeepisode of psychosis, or very early in the course of illness.Here, Weinberger, DeLisi, Perman, Targum, and Wyatt (1982) foundthat schizophreniform and chronic schizophrenic patients showedcomparable lateral ventricle enlargement, compared to controls.A second study, comparing teenage schizophrenic andschizophreniform patients to both a psychiatric and to a normalLiterature Review 34control group, found similar results (Schulz, Koller, Kishore, etal., 1983).A third report, however, failed to find the expectedresults: In the MAP Project, schizophreniform and schizophrenicpatients were compared both to medical and normal control groups(lacono, Smith, Moreau et al., 1988; Smith, 1986). Unlike thetwo other studies of early—phase patients, no significant groupdifferences in lateral ventricle size were detected. In fact,the ventricle size in the two patient and two control groups wasremarkably similar (mean VBR ±. s.d.: schizophrenics 6.7 ±.. 2.6;schizophreniform 5.9 t 2.4; normal controls 6.4 2.8; medicalcontrols 5.8 + 2.6).This inconsistency is hard to resolve. The negativefindings of Smith (1986) were not due to a lack of statisticalpower. One possible explanation relates to the nature of thesample: Many authors interpret findings in the field assuggesting that a portion (but not all) of schizophrenic patientshave LyE. The MAP Project is the first to report on a reasonablyrepresentative sample--which may include a much larger proportionof normal (small) ventricle patients than have studies which drawon hospital populations. Without a representative sample, thegood-prognosis patients who are not currently in treatment andwho are more likely to have normal—sized ventricles, areunderrepresented. If this is true, then studies that recruitpatients from treatment facilities will derive larger meanventricle sizes than studies based on representative samples.Literature Review 35The second strategy regarding the timing of ventricleenlargement examines the stability of ventricle size over time tosee if LVE is progressive. Here, three groups of investigatorshave re-examined schizophrenic patients either 2—5 or 7—9 yearsafter their initial scan. All found no change in cerebralventricular size, whether in chronic, continuously ill patients(Illowsky, Juliano, Bigelow, & Weinberger, 1988) or young, early-phase patients (Sponheim, lacono, & Beiser, 1991; Vita, Saccetti,Valvassori, & Cazzullo, 1988). Together, the results from thetwo strategies indicate that ventricular enlargement is presentamong at least some schizophrenics at onset, and is likely stableafter the first psychotic episode.Magnitude of enlargement. Several lines of evidence provideestimates of the magnitude of enlargement. The first is based onthe meta—analysis of Raz and Raz (1990), who found an effect sizeof approximately .60. Based on a normal distribution with a meanof six and a standard deviation of three VBR units, the typicalenlargement will be about 1.8 VBR units. A second approach, usedin another review, was based on raw data: Smith (1986) foundthat the median difference between the mean of schizophrenic andcontrol group ventricles was 1.7 VBR units (SD=0.7).Twin studies provide a third source of data for estimatingthe degree of enlargement. The ventricle size of normalmonozygotic (HZ) twins Is virtually identical (r=.97), andhealthy dizygotic (DZ) twins showed substantially less similarity(r=.35; Reveley et al., 1982; Suddath, Christison, Torrey,Casanova, & Weinberger, 1990). Thus, any difference in ventricleLiterature Review 36size in MZ twins discordant for schizophrenia is probably anexcellent estimate of LyE.In their magnetic resonance imaging (MRI) study, Suddath etal. (1990) found that ventricles of the affected twin wereapproximately 15-2O larger than those of the unaffected twin.In another twin study, the mean intrapair difference estimatesfor discordant and healthy MZ twin pairs were 2.16 and .36 VBRunits, respectively (Reveley et al., 1982). The differencebetween these coefficients (1.8 VBR units) can be taken as anindication of the degree of enlargement. Taken together, themeta-analysis, the estimates from raw data, and the twin studiesall derive approximately similar estimates of the magnitude ofenlargement.Prevalence. The nature of the distributions of lateralventricle size limit the ability to estimate the prevalence ofLVE. The substantial overlap in the distributions of LV size ofschizophrenics and control groups contributes to widely differingestimates of the prevalence of LV enlargement.There are several definitions of enlargement. The mostfrequently used definition is the mean plus two standarddeviations, based on a control group within the same study(Pfefferbaum, Zipursky, Lim, et al., 1988). Weinberger and hiscolleagues (Weinberger, DeLisi, Neophytides, & Wyatt, 1979) havesuggested that the cutoff be set at 10 VBR units--anything largerwould be considered an enlarged ventricle. A third method ofdefining large and small ventricles is simply to divide a sampleat the median, i.e. the “median split” technique.Literature Review 37However, in light of the absence of a bimodal distribution(Daniel & Weinberger, 1991), it is difficult to assertunequivocally that any particular cut—off point be used (Crow,1988). Psychometrically, it would appear that VBR is best usedas a continuous variable. Without a rational basis for theselection of any specific cut—off point, it seems impossible todetermine a meaningful estimate of prevalence of ventricularenlargement.There is, however, one current exception to this rule. Areasonably good estimate can be made in the case of monozygotic(HZ) twins who are discordant for schizophrenia, because theventricles of healthy HZ twins show an extraordinarily highdegree of similarity in size. Here, the combined results of twostudies of discordant HZ twins found that the affected twin hadlarger ventricles in 19 of 22 (86%) cases (Reveley et al., 1982;Suddath et al., 1990). Unfortunately, no equivalent estimatesare currently possible on the prevalence of LyE in broadersamples of schizophrenics.Correlates of Ventricle SizeIn an attempt to understand better the significance of LVEin the course of schizophrenia, research workers have studied anumber of its correlates. They include premorbid functioning,clinical status, treatment response, and prognosis.[NE and Dremorbid adjustment. If large ventricles areassociated both with poorer outcome and poorer premorbidadjustment, then any prognostic liability represented by LVE mayLiterature Review 38simply be due to deficits in premorbid ad:iustment. Thus, inorder to demonstrate any liability of large ventricles on courseand outcome, one must control for the relationship betweenventricle size and premorbid adiustment.As displayed in Table 3, most of the studies that examinedthis relationship were based on chronic patients. Two studies,however, represented a wide range of severity and functioningamong the chronic samples (Pearlson et al., 1985; Williams,Reveley, Kolokowska, Ardern, & Mandeibrote, 1985). Both foundthat large ventricles were associated with poor premorbidfunctioning, which included the presence of schizoid traits inchildhood and adolescence. Of the three studies comprised ofsamples with a restricted range of chronic patients, two detecteda relationship between poor premorbid functioning and largerventricles (Jeste, Kleinman, Potkin, Luchins, & Weinberger, 1982;Pandurangi, Bilder, Rieder, Mukherjee, & Hamer, 1988) and onereported a trend in the opposite direction (Nasrallah et al.,1983). The sixth study of schizophrenic siblings found no VBRpremorbid relationship (DeLisi, Goldin, Hamovit, et al., 1986).In summary, about two—thirds of studies of LV size andpremorbid status have documented a significant relationship,where larger ventricles have been associated with poorerpremorbid status. That LV size and premorbid status areI-. m 1Table3.VentricleSizeandPremorbidAdjustmentAuthor(s)SampleResultsmWeinbergeretal.Chronic,treatment—refractory1.Largeventriclegrouphadpoorer(1980);Jesteetschizophrenicin-patients,,b)socialfunctioning:r=.34(p.OG).DeLisietal.Schiz.siblings(26patientsNosignificantcorrelationbetweenVBR(1986)from12families),andeitherof2premorbidscores.Nasrallahetal.Chronicschizophrenicmen.Authorssuggestnopremorbiddifferences(1983)inVBRgroups,althoughre—analysissuggestsotherwise(seetext).w ‘.0Literature Review 40associated is consistent with the notion that ventricularenlargement occurs prior to the onset of schizophrenia, but onlyin some portion of afflicted people. An alternative explanationis that premorbid status is a mediating variable in etiology,e.g., people with poor premorbid functioning are more prone toventricular enlargement. Either way, the implication forresearch projects that attempt to predict course and outcome isthat premorbid adjustment should be included as a controlvariable.Severity of illness. In order to interpret a relationshipbetween ventricle size and prognosis as possibly causal, one mustalso assess any concurrent presence of [NE with overall severityof illness. After all, it may be that VBR simply reflects a moresevere form of illness.Unfortunately, data are sparse. DeLisi et al. (1986)examined the relationship between VBR and a global measure ofseverity. Their sample consisted of 26 schizophrenic siblingsfrom 12 families; other sample characteristics were not reported.The results from this study revealed a nonsignificant trendbetween VBR and an overall severity rating (r=.33, p<.1O). Again,the lack of description of the sample makes the findingsdifficult to interpret.In a meta-analysis, Raz and Raz (1990) assessed thehypothesis that severe forms of illness would have largerventricles. After controlling for patients’ age and length ofillness, the cumulative length of hospitalization accounted for21% of the variance (partial =.46) in the effect size ofLiterature Review 41studies that compared the size of lateral ventricles inschizophrenics and control groups. In other words, severity ofillness is associated with lateral ventricle size.In pursuing the severity-VBR association, much of the work hasbeen directed at two more specific correlates: Thepositive/negative syndromes, and treatment response. Regardingthe former, several overlapping but distinct concepts have beenproposed, including the Type I vs. II distinction of Crow (1980),and the deficit syndrome of Carpenter and Strauss (e.g.,Carpenter, Heinrichs, & Wagman, 1988). More recently, Liddle(1987) has described three factors: positive symptoms, adisorganization cluster, and the usual negative symptom*groupingIn specifying the nature of the severity, most investigatorshave found hypothesized correlations between large VBR and anincreased number of negative symptoms (Andreasen, Olsen, Dennert,et al., 1982; Kemali, Mai, Galderisi, et al., 1985, 1987; Owens,Johnstone, Crow, et al., 1985; Pearison, Garbacz, Breakey, Ahn, &DePaulo, 1984; Takahashi, manage, Kato, et al., 1981; Williamset al., 1985), although several studies have failed to detectsuch an association (Nasrallah, Kuperman, Hamra, & McKalley—Whitters, 1983; Pandurangi et al., 1988). The relationshipbetween increased ventricle size and negative symptoms has alsobeen noted in most MRI studies (e.g., Andreasen, Nasrallah, Dunn,* Whatever the particular concept, it is important to recall thatthese are clusters of characteristics, akin to the notion of‘fuzzy sets’, rather than discrete subgroups of patients who eachhave all of the characteristics.Literature Review 42et al., 1986; Andreasen, Erhardt, Swayze, et al., 1990; Besson,Besson, Cherryman, et al, 1987; Olsen, Nasrallah, Coffman, &Scwartzkopf, 1990; although not in Mathew, Parein, Prakesh, etal., 1985).Treatment response. If negative symptoms and large ventriclesare indeed associated, this begins to resemble Crow’s Type IIsubtype, where one characteristic is poor response to medication.A number of studies have prospectively assessed VBR and responseto neuroleptic medication. In general, prospective studies havefound that patients with large ventricles show poorer treatmentresponse (Jeste et al., 1982; Luchins, Lewine, & Meltzer, 1984;Smith & Maser, 1983; Weinberger, Cannon—Spoor, Potkin, & Wyatt,1980). Retrospective investigations, on the other hand, havefound no relationship between VBR and medication response (e.g.,Nasrallah et al., 1983; Williams et al., 1985).LV size and ronosis in schizophrenia. Several approacheshave been taken to attempt to link lateral ventricle size withgeneral prognosis in schizophrenia, rather than that suggested bytreatment response over a number of weeks. Since this issuebears directly on one of the aims of this thesis, the evidencehere will be examined in some detail. As will be seen below(Table 4), the results from the studies reported to date show amodest but consistent relationship.One study found an equivocal relationship between VBR andoutcome among three groups of schizophrenics selected torepresent a good, poor, or intermediate course of illnesst-II-,. 1Table4.VentricleSizeandPrognosisAuthor(s)SampleResultsDeLisietal.First—episodeschizo—1.Nodifferencein2—yearoutcomebetween(1983)phreniformpatients.largelargeVBRpatientsandallothers.2.Extremegroups:thelargestVBRgrouphadlowerscoreson2outcomemeasures.Pearlsonetal.Schiz.patients,eithercontin—MeanVBR:unemployedpatients>(1984)uouslyemployedorunemployedemployedpatients,controls.vs.matchednormalcontrols.Williamsetal.Schiz.patients:eithergood,Comparingextremegroups,largeVBRmore(1985)intermediate,orpooroutcome,prevalentintheworstoutcomegroup.Luchins&Chronicschiz.pta.fromacute-MeanVBR:chronicwardpatients>l4eltzer(1986)careorcontinuous—careward.acutewardpatients.Katsanisetal.First—episodeschizophrenicsUsingmediansplit,largeVBRgrouphas(1991)fromMAPProject.pooreroutcomeatboth9and18months.LA)Literature Review 44(Williams et al., 1985). When VBR was used as a continuousdependent variable, an analysis of variance (ANOVA) among thethree outcome groups was not significant. However, when VBR wasdichotomized and used as a grouping variable among the good-outcome subset of cases, significant results were obtained--therewere more patients in the good—outcome group with normalventricles, than with large ventricles.Luchins and Meltzer (1986) used a cross—sectional strategyin comparing two groups of chronic schizophrenic patients, onefrom an acute-care ward and the other from a continuous-careward, which were matched for total duration of illness. Theirresults showed that the mean VBR was higher and a greater portionof large ventricles were observed in the continuous-care, poor-outcome group.Pearison et al. (1984) assessed the association VBR andemployment as a measure of outcome. The authors found that thepersistently unemployed patients had significantly largerventricles, compared to their continuously employed counterparts.Furthermore, the mean ventricle size of employed patient groupswas strikingly similar to that of a normal control group.Two studies investigated prospectively the relationshipbetween ventricle size and outcome among first—episode samples.The first (DeLisi, Schwartz, Targum, et al., 1983) divided theirsample into normal- and large-ventricle groups: They found amodest but nonsignificant trend for normal-ventricle patients toshow better outcome. Stronger and significant results wereobtained when the best and worst outcome quartiles wereLiterature Review 45contrasted: People with large ventricles were overrepresented inthe worst outcome group. In the other study of a first-episodesample, schizophrenics with larger ventricles had significantlypoorer short—term outcome (Katsanis, lacono, & Beiser, 1991). At18-month follow-up, MAP Pro.iect patients with larger ventriclesize had higher symptom checklist scores. Nonsignificant trendsfor large VBR patients to have poorer work functioning and poorerglobal rating scores were also apparent. These findings weremaintained even after controlling for the level of premorbidfunctioning. Moreover, the results were true for theschizophrenic group but not for the affective psychosis group.These results, if replicated, suggest that the pathophysiologicprocess(es) responsible for LVE may have a modest effect onshort-term course and outcome, over and above any effect theymay have had on childhood and adolescent adaptive functioning.In sum, every study that has addressed the issue has found arelationship between larger ventricles and poorer outcome inschizophrenia, albeit two of the six studies reported mixed(positive and null) findings. The only study that has controlledfor obvious confounding factors is from the MAP Proiect team,where LV size was associated with 18-month outcome, aftercontrolling for premorbid status. Together, these findings pointto an association between LV size and prognosis that is modestbut robust. When the studies are limited to those assessing afull range of patients, the findings are limited to the shortterm course of schizophrenia. Whether the relationship continuesto be true for longer-term outcome remains to be seen.Literature Review 46Ventricle Size and Affective DisordersIn CT studies of brain structures, people with Bipolar andMajor Depressive Disorders have served as psychiatric comparisongroups. If ventricular enlargement is present in affectivedisorders, it may reflect a similar, nonspecific etiologicalfactor as is present in schizophrenia. The hazard may beindependent and additive to the etiology of affective disorders,it may be etiologically related, or it may be an epiphenomenonsecondary to major psychopathology in general.Table 5 summarizes the published reports of theinvestigation of LyE in affective disorders, grouped according toBipolar, Depressed or those that have studied both types ofaffective patients.The VBR of Bipolar patients has been the subject of sixcomparisons: As shown in Table 5, three of those studies (Dolan,Calloway, & Mann, 1985; Nasrallah, McCalley-Whitters, & 3acoby,1982; Pearlson et al., 1984) reported finding larger ventriclesin the bipolar patients, compared to nonpsychiatric controls. Afourth (Dewan, Haldipur, Lane, et al., 1988) found an excessprevalence of large ventricles, although no difference betweenpatients and controls in mean VBR. Of the three studies (Dewanet al., 1988; Nasrallah et al., 1982; Pearison et al., 1984) thatprovided prevalence data, all noted that about one-Table5.VentricleSizeandAffectiveDisordershAuthor(s)SampleResultsNasrallahetal.Bipolarvs.schiz.pts.1.MeanVBR:bipolar,schlz.>controls.(1982)vs.medicalcontrols.2.LargeVBR:29%ofbipolars;34%schiz.mRiederetal.Bipolarvs.schiz.vs.MeanVBR:nogroupdifferences.(1983)schlzoaffectivepts.1-..Pearlsonetal.Bipolarvs.schiz.pts.1.MeanVBR:a)schiz.,bipolar>controls;(1985)vs.normalcontrols.b)continuouslyunemployed>employed.2.VBRunrelatedtonumberofhospzns.Dewanetal.Bipolarpatients,vs.1.MeanVBR:nogroupdifferences.(1988)medicalcontrols.2.%withlargeventricles:bipolar>schiz.3.Severity:nodifference(onanyindex)betweenlargeandnormalVBR.Dolanetal.Bipolarvs.depressive1.MeanVBR:bipolar=depr.pts.>controls.(1985)pts.vs.normalcontrols.2.VBRunrelatedtooutcome.laconoetal.Schiz.,schizophreniform,1.MeanVBR:nogroupdifferences(1983)bipolar,&depr.pta.vs.2.Large/smallVBRunrelatedtooutcomemedical&normalcontrols.forbothaffectivegroups.Roy—ByrneBipolarvs.depressedpts.MeanVBR:Bipolarsdepressives.etal.(1988)Jacoby&LevyElderlydepressivesvs.1.MeanVBR:nogroupdifference.(1980)matchedhealthycontrols.2.Outcome:largeVBRpts=normalVBRptaScottetal.Severelydepressedpta.MeanVBR:Depr.pta.>medicalcontrols.(1983)vs.normalcontrols.Targumetal.Delusionalvs.nondeludedMeanVBR:Delusionaldepressives>non(1983)depressivesvs.controlsdelusionaldepressives,controls.Shimaetal.Depressedpta.vs.MeanVBR;a)allpatients)controls;(1984>medicalcontrols.b)poor—outcomepta>good-outcomepts.Literature Review 48third (range 29-35%) of bipolar patients had VBR scores thatexceeded the control group mean by two standard deviations.Taken together, the two indices of LyE, mean VBR and prevalenceof large ventricles, suggest that a substantial minority ofbipolar patients do have enlarged ventricles.Among people with Major Depression, the evidence regardinglarger ventricles (compared to nonpsychiatric controls) isinconclusive. Three of the studies (Dolan et al., 1985; Scott,Golden, Ruedrich, & Bishop, 1983; Shima, Shikano, Kitamura, etal., 1984) reported positive results, and a fourth found positiveresults for delusional but not nondelusional depressives (Targum,Rosen, DeLisi, Weinberger, & Citrim, 1983). Two others found nodepressive-control differences (lacono et al., 1988; Jacoby &Levy, 1980). Three of the studies assessed prevalence: Jacobyand Levy (1980) found that 29% of their elderly sample had largeventricles, and Targum et al. (1983) reported a similar figure(25%) among their delusional patients. None of theirnondelusional depressives had large ventricles.The conclusion to be drawn from the aggregated studies ofaffective disorder patients to date is similar to that based onstudies of schizophrenics: A majority, but not all, of theinvestigations have found evidence of LVE in both Bipolar andMajor Depressive disorders, using group mean VBR or the increasedprevalence of large ventricles. Precise estimates of prevalence,or descriptions of the subgroup of patients who can be expectedto exhibit LyE, are currently not possible.Literature Review 49Correlates of ventricle size in affective disorder. Researchexamining the correlates of ventricle size among people withaffective disorders is limited. To date, a small number ofpapers have examined correlates such as premorbid status andprognosis.Pearison et al. (1985) compared premorbid status amongsubgroups of bipolar patients who did and did not exhibitventriculomegaly. Their global index assessed the entirepremorbid lifespan, from perinatal complications to cumulativeemployment history. The results indicated that bipolar patientswith and without large ventricles were not different with respectto premorbid status. Similar null results were reported by Dewanet al. (1988), who used a different global rating to assesspremorbid functioning in their Bipolar sample.Four studies have attempted to relate LVE to prognosis inaffective patients. In one study of intermediate— or good-outcome bipolar patients, the results indicated an equivocalrelationship between ventricle size and outcome (Pearlson et al.,1984). When treated as a continuous variable, VBR showed norelationship to two measures of outcome. When the authorsdivided the bipolar patients into quartiles based on VBR scores,and contrasted the extreme groups, however, the largest—ventriclegroup had poorer outcome on both measures than did the group withthe smallest ventricles. Another study of depressed patientsfound that those with poor short-term outcome had larger mean VBRcompared to both good—outcome patients and controls, whose scoresLiterature Review 50did not differ (Shima et al., 1984); these differences wereapparent, however, only among late—onset (>50 years) patients.Two studies reported unequivocal negative results. Jacobyand Levy (1980) failed to find an association between short-termoutcome and ventricle size among elderly depressed patients. Thefinal study of this type is the report by Katsanis et al. (1991)on the MAP Proiect samples. As described above, they relatedthe LV size of unipolar and bipolar patients with psychoticfeatures to three measures of outcome, while controlling forpremorbid status. Three measures of outcome, taken at both 9—and 18-month follow—up, were not related to ventricle size ineither affective group.In sum, neither of two reports dealing with premorbid statusamong bipolar patients found an association with ventricle size.Regarding the association between outcome and ventricle size, theresults are mixed but largely negative. Among depressedpatients, the conclusion regarding a VBR-outcome association issimilarly negative. Overall, the weight of evidence suggeststhat, while lateral ventricular enlargement is likely presentamong affective patients, there is no relationship betweenventricle size and either premorbid status or prognosis.Summary: VBR as a Predictor of OutcomeIn this section, we have seen that LV size is a good, ifimperfect, index of structural changes in the schizophreniadisease process. Most, but not all, studies found thatschizophrenic people have larger ventricles than medical orLiterature Review 51normal controls. The magnitude of the group difference isequivalent to 1.8 VBR units, or approximately 30% of originalsize. The timing of the enlargement is probably at or beforeonset of the disorder, and further changes are unlikely.The crucial question of the prevalence of enlargement remainsunanswered. The distribution of ventricle size in schizophrenicpeople is smooth; it is not bimodal, as it would be if somepeople had a substantial degree of enlargement and others hadnone. In the special case of discordant MZ twins, almost all ofthe probands have ventricles larger than the unaffected twin.The most likely explanation is that enlargement is itselfcontinuously distributed--some patients have a small increase inLV size, most have a modest degree of enlargement, and some havestrikingly larger ventricles.Of more direct relevance to this dissertation project are thecorrelates of large ventricles, which include poor premorbidfunctioning, severity of illness (particularly an excess ofnegative symptoms), and poor response to neuroleptic medication.In this light, the correlate of poor prognosis is meaningful:Most relevant to this thesis is that all of the previous studieshave found either an equivocal or an unqualified significantrelationship between LV size and prognosis. Of these, the mostrelevant is the paper describing short-term outcome of MAPProject participants (Katsanis et al., 1991), where ventriclesize was significantly related to three measures of short-termoutcome, after controlling for premorbid functioning.Literature Review 52Previous reports that have assessed ventricle size andoutcome in affective disorder have been nearly unanimous infailing to detect a relationship between the two variables.While most studies have detected group differences in LV sizebetween affective patients and nonpatient controls, the increasedventricle size does not seem to have any apparent significancefor outcome, nor for premorbid functioning.While there seems to be substantial empirical support for thehypothesis that LV size will predict outcome, the VBR-outcomerelationship will likely be of moderate size. The principalreason relates to the distinction between ventricular size andenlargement: The median extent of enlargement is small (1.8 VBRunits), compared to the wide range of variability inschizophrenic and control samples (4—11 VBR units). This meansthat ventricle size--even if measured at the onset of thedisorder——is not a powerful strategy for assessing the presenceor prevalence of enlargement. More importantly, the magnitude ofcorrelations between outcome and ventricle size will be muchsmaller than between outcome and enlargement per se.Unfortunately, no measure of enlargement is currently available.Literature Review 53Biological Factors: II. Smooth Pursuit Eve MovementsIf the research on brain structural attributes inschizophrenia has been empirically driven, due to theavailability of the new imaging technology, then another avenueof investigation-—that of impaired smooth pursuit eye movements-—has been theoretically driven. Because the research has beenguided by the conceptual requirements of genetic (trait) markers,the literature is more coherent, and the implications of thefindings are clearer.This section is organized in a manner similar to precedingsections. First, I will present brief introductions to themechanics of oculomotion, and to the requirements of the theoryof genetic markers. Second, I will assess the putative trait ofimpaired oculomotion in schizophrenia vis—a-vis these formalrequirements. In the third section, I propose impaired smoothpursuit eye movements (SPEM) as an index of biologicalvulnerability. To the extent that this argument is persuasive, Iwill finally review the literature consistent with the notion,viz, the correlates of impaired eye movements.Systems Governing Eve MovementsThere are three principal oculomotor systems, each with itsown function and anatomy. The first is the smooth pursuit eyemovement system, which is concerned with keeping a moving objectof interest on the fovea of the retina. The SPEM system isdesigned to track the kind of motion that is illustrated by aLiterature Review 54smoothly oscillating pendulum, i.e. relatively slow (less thanforty degrees of arc per second) and mostly lateral movements. Asecond oculomotor system governs saccades, relatively fast eyemovements, wherein vision during the change of eye position isunimportant. Saccades provide for quick changes in eye position,and also serve to correct errors generated in the SPEM tracking.The third visual system is the vestibulo-ocular reflex (VOR),which is used by a moving person to keep a stationary target insight.Dysfunction in the SPEM system is not specific toschizophrenia. It is present in disorders such as alcoholism andmultiple sclerosis, which are associated with diffuseneuropathology. SPEM impairment is also noted in localizeddisorders, like Parkinsonism (Johnston & Pirozzolo, 1988).It is difficult to infer the neuroanatomical significance ofabnormal SPEM, because the control systems for eye tracking arediffusely located in the human brain. Failure of the smoothpursuit system can be caused by a failure in a number of corticalareas, e.g. frontal eye fields, posterior, parietal, or middletemporal regions. Similarly, it can be a result of failure insubcortical areas associated with the modulation of eyemovements, such as the basal ganglia, cerebellum, or pons(Thaker, Buchanan, Kirkpatrick, & Tamminga, 1989). Smoothpursuit, therefore, is only a general indicator of oculomotiondysfunction: it lacks anatomic specificity.Literature Review 55Recording and Scoring SPEMIn the early studies of SPEM, subjects visually tracked asmall weight that was attached to a string. While this methodwas easy to use, it did not invite a lot of precision. A moreprecise instrument is an oscilloscope, where a dot moves back andforth on a screen tracing approximately twenty degrees of visualarc.When movements of the stimulus are recorded on moving graphpaper, a smooth sine wave is produced. If eye-trackingperformance is good, the movement of the eyes will replicate thatsmooth sine wave. When tracking is poor, however, the tracing ofthe eye movements will be jerky and irregular.To measure lateral eye movements, two techniques have beenwidely used. The first is infrared reflection, where a lightsource is mounted near the eye. Because the pupil and the irisare not equal in their ability to reflect light, and becausethere is a sharp boundary between the two, photocells placed atthe outer edge of the eye can measure reliably differences in theamount of reflected light. The ratio of more to less reflectedlight at various locations around the eye is the basis fordetermining eye position at any point in time.An alternative is the measurement of small electricalpotentials generated from the eye itself. Within the eyeball, asmall standing electrical charge is always present: the retina,because of the high density of neurons, generates about onemillivolt of potential more than the cornea. If two electrodesare placed at the outsides of the eyes (the canthi), and the eyesLiterature Review 56are looking straight ahead, there will be no differences betweenelectrodes in the amount of current measured. When the eyes movelaterally, the retinas, with their greater electrical charge,move closer to one electrode. The small difference in current,relative to baseline (eyes forward), is the index of thedeflection of the eyes.After sources of artifact such as eye blink or head movementare eliminated, the data are subject to either global ratings orto mathematical scoring. The most frequent global rating systemis a dichotomous one (Holzman, Proctor, & Hughes, 1973) of goodor poor tracking performance, although a five—point scale hasalso been used (Shagass, Roemer, & Amadeo, 1976). Mathematicalscoring has been done using a gain score or signal—to—noisemeasure (Lindsay, Holzman, Haberman, & Yasillo, 1978), but ismore often subject to a root-mean—square (RMS) error analysis.Here, the actual tracking performance--after correcting for phaseand amplitude differences-—is superimposed on to that generatedby the stimulus, and then the average distance between the twotracings is computed. All of these measures, whether globallydefined or mathematically computed, correlate well with eachother. The global method of scoring is less sensitive than themathematically-scored approaches, and it may over-estimate theprevalence of SPEll dysfunction (Clementz, Grove, lacono, &Sweeney, 1992).Literature Review 57Theoretical Requirements for a Genetic MarkerTo satisfy the requirements for a putative genetic marker, acharacteristic must meet four general requirements. It shouldhave trait properties, i.e. it should be stable over time, and beunaffected by clinical state. It must show relative specificityand sensitivity to a disorder. It should also show increasedprevalence in unaffected family members. Finally, the proposedmarker should segregate with the disorder in family members. Aswill be seen below, impairment in smooth pursuit eye movementslargely fulfills those theoretical requirements.Until recently, the prevalence of impaired SPEM was estimatedto be 60-80% among schizophrenic people, compared to 8% amongnormal control subjects (Holzman, 1992). More recently, however,estimates of prevalence derived from mathematically-scored datafrom three independent samples have generated much lowerprevalence estimates (20-37% of schizophrenic patients; Clementzet al., 1992). SPEM dysfunction is present in mood disorders,but only as a state measure: it is not present, for example,among people diagnosed as Major Depression in Remission(Szymanski, Kane, & Lieberman, 1991). Whether most or only asubstantial minority of schizophrenic patients have impairedSPEM, does not alter the fact that SPEM dysfunction shows good(relative) specificity to schizophrenia. Hence, it meets thiscriterion for a genetic marker.Until recently, it also appeared that the related criterionof high sensitivity had been met (prevalence 51-85% of diagnosedcases; Holzman, Kringlen, Matthyse, et al., 1988). Using theLiterature Review 58mathematical scoring procedures, relative sensitivity is stillachieved, with 20—37% of schizophrenic patients so identified.However, if the unit of analysis for sensitivity is the family ofa schizophrenic proband, then the sensitivity is approximately.75 (Clementz et al., 1992). Thus, this requirement of a geneticmarker is substantially fulfilled.SPEll dysfunction in schizophrenics has good traitproperties, since it is present among both remitted andsymptomatic patients (lacono, 1988). It has been shown to bestable over periods as long as two years (lacono & Lykken, 1978).SPEM functioning is unaffected by most medication: trackingperformance by patients during drug—free periods is much the sameas while on antipsychotic medication. SPEM function is affectedby lithium carbonate and several CNS depressant medications, butthese reflect state (rather than trait) variations. Finally,SPEM dysfunction in schizophrenic people does not appear to be afunction of an inability to maintain voluntary attention,distractibility, a lack of motivation, or a generalized deficit(Szymanski et al., 1991).Another requirement for a genetic marker is that it showincreased prevalence among family members, and in particular showevidence for substantial heritability. Among studies of twins,the results suggest that SPEM shows a substantial degree ofheritability: the intraclass correlations derived from healthymonozygotic twins average 0.68, whereas those from healthydizygotic twins average 0.35 (lacono, 1988). Similarly, impairedSPEM shows increased prevalence among parents and adolescentLiterature Review 59children of schizophrenic patients, compared to similar relativesof healthy control samples. Using global ratings of eye-tracking, between one-third and one-half of the first-degreerelatives of schizophrenics showed SPEM dysfunction, compared toabout 10% of the relatives of mood disorder patients (Holzman,Soloman, Levin, & Waternaux, 1984). Using RMS error or gain—score approaches, eye-tracking dysfunction is present among 19-22% of first—degree family members in schizophrenia.The presence of schizotypal or schizoid personalitytraits/disorder is often taken to represent the presence of aschizophrenic genotype. If impaired SPEM is indeed associatedwith a genotype, then poor tracking should be significantlyassociated with schizoid or schizotypal traits in nonpsychiatricsamples. This was the reasoning used by Siever, Coursey,Alterman, et al. (1984), who screened 284 male college studentsfor eye—tracking abilities. They administered a diagnosticinterview to the best and the worst 109o of eye-trackers in thesample. They found that a significantly greater number of poor-tracking students (54%) met DSM—III criteria for schizotypalpersonality disorder, compared to 11% of the good-tracking group.One final theoretical requirement of a genetic marker is thatit should segregate with the disorder, e.g. good-trackingschizophrenics should have good-tracking (nonschizophrenic)relatives. Some investigators (Holzman et al., 1984; Matthyse,Holzman, & Lange, 1986) have provided evidence, however, of goodtracking patients with poor-tracking but otherwise healthyrelatives. The authors explained their findings with the conceptLiterature Review 60of a latent trait or pleiotropy: the genotype can be expressedas poor SPEM, as schizophrenia, or both.Correlates of SPEM DysfunctionImpaired oculomotion is often considered an impairment in theability to sustain involuntary attention. Since smooth pursuitimpairment is present before and after a psychotic episode, then(in)ability to maintain basic attention would likely influenceinter-episodic functioning. To propose it as a predictor ofcourse and outcome, one would want to assess several of itscorrelates. For example, in order to understand its unique andindependent status as a predictor, one would wish to know thedegree to which it co-occurs with other putative indices ofbiological liability. Further, if the putative genetic marker isindeed an index of vulnerability, it should be relatedindependently to premorbid status and clinical characteristics.Finally, the vulnerability marker may be related to prognosis.SPEM and lateral ventricular enlargement. The researchrelating SPEM dysfunction to other indices of vulnerability hasbeen relatively sparse. Of the papers that have been published,it is the relationship between SPEM and ventricle size that hasbeen the most frequent object of study.In a study of chronic, treatment-refractory patients,markedly disordered eye tracking was disproportionately prevalentamong patients with large ventricles (64%), compared to thosewith normal ventricles (30%; Weinberger & Wyatt, 1982). Inanother study where patients represented a range of severity, aLiterature Review 61nonsignificant correlation of r=.35 between eye movementdysfunction and LV size was detected (Bartfai, Levander, Nybeck,Berggren, & Schalling, 1985). The correlation may have beenattenuated due to the use of linear measures of ventricle size,and statistical power was limited due to the small sample size(n=18). In a third study assessing ventricle size and eyemovements, correlations of r=.28—.35 were reported (Siever, vanKammen, Linnoila, et al., 1986). Although these coefficientswere not statistically significant, the sample size was small(n=13)The only report of the relationship between ventricle sizeand SPEM to use relatively large and representative samples comesfrom Smith (1986), who reported on three diagnostic groups fromthe MAP Project. Pearson correlations between VBR and SPEM forschizophrenic and schizophreniform groups were trivial (r=.02 andr=-.13, respectively).Taken in the aggregate, these studies show a weakrelationship between the two variables—-if one exists at all.The magnitude of the relationship can be expected to be increasedif assessed among patients with more severe forms of illness,since both SPEM and VBR have been independently associated withseverity. Such an interpretation is possible in three of thestudies (Bartfai et al., 1985; Siever et al., 1986; Weinberger &Wyatt, 1982), where coefficients in the order of 0.3 wereobserved. Only in the Smith (1986) report was the associationbased on a more representative sample. There, the correlationswere close to zero for the schizophreniform and schizophrenicLiterature Review 62patients. Thus, the true magnitude of association may be nearzero. This suggests that the two indices of biologicalvulnerability are virtually independent, and may represent twodistinct pathophysiological processes.SPEll and premorbid ad1ustment. If SPEll dysfunction indexes adisorder of nonvoluntary attention, and if it is a trait that isunrelated to symptom status, then one might expect it to bepresent prior to the onset of schizophrenia. Only two studieshave directly assessed the relationship between impaired SPEll andpremorbid adjustment. In one small study (n=22), there was nosignificant relationship between eye tracking and adolescentsocial adjustment (Bartfai et al., 1985). The actual coefficientwas not provided.In another small study (n=14), premorbid social adjustmentwas correlated both with qualitative and quantitative ratings ofSPEll. Neither of the correlations achieved significance(qualitative measure: r=.48, p<.lO; quantitative measure r=.31,p=n.s.; Siever, van Kammen, Linnoila, Alterman, Hare, & Murray,1986).The small sample size in each of these studies greatlyhampers any interpretation of the aggregate results. The onlyconclusion that can safely be drawn is that, if a correlationbetween SPEll dysfunction and premorbid adjustment exists, it isnot of an overwhelming magnitude.SPEll and clinical characteristics. In the literature, thereis a scattering of results in the area of the clinical correlatesof SPEll. Siever et al. (1986) speculated that poor eye trackersLiterature Review 63may represent a more severe subgroup of schizophrenia. Inassessing a small sample of chronic, treatment—refractoryschizophrenic patients, they found no significant relationshipbetween severity of psychosis and SPEM (r=.17). The lowreliability of their single-item measure of severity, however,may have contributed to the negative results.In another study, Bartfai et al. (1985) used apsychometrically superior measure of global severity of illness.Despite this improvement, there was no relation between SPEM andthe global severity rating. Thus, if one can draw tentativeconclusions from these two studies, it may be that eye-trackingdysfunction does not imply a general liability such as a moresevere type of illness.SPEM and prognosis. If SPEM represents a specific liabilityindependent of others, it may be reflected in course and outcome.Again, only two studies have addressed this issue. Siever et al.(1986), in their study of 14 young patients, did not find asignificant relationship between the total number of lifetimepsychiatric hospitalizations and SPEM measures (r=—.31, withpoorer trackers having more admissions). Once again, the smallsample size limits the interpretation of the negative findings.The only other reference is in an edited chapter by lacono(1988). There he briefly described previously unpublished datafrom the MAP Project, based on 56 schizophrenic participants. Asignificant relationship between SPEM and 18—month outcome waspresent, with 90% of poor trackers having a poor outcome,compared to only 30% of good trackers. No definition of poorLiterature Review 64outcome was provided in that report, nor has a more detailedversion of the results been published. lacono concluded thatbiological vulnerability may play a contributing role not only inthe etiology, but also the maintenance of schizophrenia.SummaryThe literature regarding smooth pursuit eye movements isrelatively tidy, even if the precise mechanism that isdysfunctional in schizophrenia is not understood. We have seenthat the theoretical requirements for a genetic marker have beenlargely met. If one accepts the Holzman and Matthyse (Holzman etal., 1988) hypothesis of a latent trait, however, then SPEMimpairment stands well as a genetic marker for schizophrenia.Together with other, possible genetic markers, SPEM mayeventually serve as a means to ‘triangulate’ ultimategenetically—based etiological factors.Until such a time, SPEM impairment may be used as an index ofone kind of biological vulnerability, which may useful inpredicting the course of the disorder. We have seen that littleresearch has been dedicated to this purpose: From the literaturereviewed above, SPEM impairment is largely or completelyindependent of another leading candidate for biologicalvulnerability, i.e. enlarged ventricles.If SPEM impairment is indeed a biological liability, thenearly signs may be evident. Using the level of premorbid socialadjustment, two studies (each with small samples) detectedcorrelations between SPEM and premorbid adjustment that were inLiterature Review 65the expected direction, even though each was statisticallynonsignificant. Similarly, if SPEM is indeed a biologicalliability, it may presage a poorer course. This was the case inone brief report from the MAP Project regarding short—termoutcome (lacono et al., 1988), and a nonsignificant trend wasnoted in a small study reported by Siever et al. (1986).Overall, there may be good reason to hypothesize that SPEMimpairment will be associated with a poorer prognosis. Thisproject is the first to assess that relationship on a largesample, and with a wide range of patients. This project is alsothe first to assess the unique and independent contribution ofSPEM impairment on long—term outcome, while controlling for otherfactors that may simultaneously affect the course ofschizophrenia.Literature Review 66General Issues in the Prediction of Course and OutcomeThe tradition of identifying predictors of the course ofschizophrenia has a long and illustrious history. Of mostinterest have been predictors that are evident prior to or at thetime of the first episode, both because of the notion that theyare untainted by the effects of lengthy illness, and because ofthe implications for early and presumably more effectiveintervention. For more extensive reviews of predictors ofoutcome, interested readers may consult Stoffelmayr, Dillavou,and Hunter (1983) or Beiser and lacono (1990). The variablesgermane to the proposed study, sociodemographic factors, aspectsof premorbid functioning, and characteristics of the mode ofonset, are briefly reviewed below.Sociodemographic and Premorbid FactorsDemographic variables such as age at onset, marital status,and sex have been repeatedly associated with outcome. Forexample, an early age at onset is associated with a pooreroutcome (=.15—.25), suggesting that younger people have a moreserious form of illness, or that they do not have a chance todevelop their social skills, occupational training, or even basiccognitive capacities (Bromet, Harrow, & Kasi, 1974; Burstein,Adams, & Chapman, 1974).Marital status is slightly stronger as a predictor (L=.2l—.32), with people who have ever been married having a betterprognosis, compared to those who have not. Marital status as aLiterature Review67predictor is usually seen as reflecting better premorbid socialskills and a later age of onset (Lorei & Guel, 1973; Stephens,1978). Further, marital status among people with schizophreniais strongly related to sex, where women tend to be married orhave been married by the time of illness onset.Sex is also related in other ways to outcome. Compared tomen, women tend to be have briefer hospital stays and manage tostay out of hospital for longer periods of time (Angermeyer,Kuehn, & Goldstein, 1989). They tend to have higher levels ofpremorbid functioning and later onset (Goldstein, 1988).Further, the superior outcome of women over men appears to berelated to their greater representation among familial types ofschizophrenia, where the nonfamilial cases tend to have a pooreroutcome (Goldstein, Santangelo, Simpson, & Tsuang, 1990).Finally, schizophrenic women may have an advantage due to therelative dopamine—inhibiting effects of estrogens (Seeman, 1982).Thus, in assessing the role of novel predictors of outcome, onemust account for several overlapping predictors that are alreadywell-known.Duration of Onset and the Course of SchizophreniaIn order to assess the importance of any baseline variable(measured at the onset of the disorder) on the outcome ofschizophrenia, one must control for the effects of varyinglengths of onset. For example, any differences in socialrelationships observed at intake may simply be the result of aLiterature Review68longer and insidious onset and as such may be a feature of thedisorder, and not a premorbid characteristic.The notion that acute (vs. insidious) onset predicts goodprognosis in schizophrenia has achieved legendary status as aprognostic indicator (cf. Bland, 1982; Docherty, van Kammen,Sins, & Marder, 1978; Neale & Oltmanns, 1980; Strauss &Carpenter, 1977). This conclusion may be premature, because thearea suffers from a number of major limitations.The first limitation relates to the failure to clearlyconceptualize——and hence assess——onset per se. Of the sixempirical reports, illustrated in Table 6, only one has used theconcept of onset to characterize the illness itself. Vaillant(1964) used as a definition the duration from the first signs ofthe episode to the appearance of florid psychotic symptoms, i.e.the prodromal period. All others have used the concept of onsetto describe the time lag between the appearance of psychoticsymptoms and treatment contact (usually admission to hospital).A second limitation is that the studies that haveinvestigated the concept appear to be based, with one exception,on chronic or mixed groups of patients. As such, the predictorin question is not the onset of illness, but the onset ofsubsequent episodes. Inasmuch as the characteristics of laterepisodes are influenced by factors such as treatment andmedication, the predictive role of later episodes is probablyunlike that of the first episode, i.e. the onset of the illness.Moreover, the use of readmitted, chronic patients probablyTable6.OnsetCharacteristicsandOutcomeinSchizophreniaAuthor(s)SampleResults1Stephens&AstrupRetrospectivefollow—upof1.Reactivesubtypehasbetteroutcome.(1963)patientshospitalized2.Characteristicsrelatedtobetteroutcomebetween1944—1954.includeacute(vs.insidious)onset.Vaillant(1964)ConsecutiveadmissionsAcuteonsetmorefrequentamongpatientsin1947-50and1961—62.whoachievedfullremission(82%)thanamongthosewhodidnot(42%).Cancro&Consecutivemaleadmissions.‘Dorrelationbetweenmodeofonsetand:Sugerman(1968)a)9mo.hosp.status:r=.21(p<.O5)b)5yr.hosp.status:r=.14(n.s.)c)Daysinhospital:r=.22(p<.O5).Bursteinetal.Mixedgroupofmales;first-Modeofonsetunrelatedtoin/outof(1974)andmultiple—admissions.hospitalatfollow—up,ordaysinhospBrometetal.In—patients.Correlationbetweenmodeofonsetand:(1974)a)hosp.daysoroccuplfunctg:n.s.b)socialadjustmentandglobalratingbothr>.35(p<.O5).Stephensetal.Retrospectivefollow-upofAcuteonsetcorrelatedwith(1978)pts.firstadmitted1948—59.outcome(r=.24,p<.05)MacMillanetal.Prospective2—yearfollow—upShorterdurationofonset(<1wk.and(1986)offirst—episodepatients1-4wks.vs.>4wks.)relatedtoreducedprobabilityofrelapse.0-ILiterature Review 7reduces the variance in outcome and thus any association with apredictor variable.A third, psychometric limitation is that the studies of onsethave tended to dichotomize a continuous variable: mostinvestigators used the simple dichotomy of acute/insidious, basedon a six-month cut_off*. Only one group of investigators, itseems, has gone beyond the dichotomy. As shown in Table 6,MacMillan, Crow, Johnson, & Johnstone (1986) described onset intheir sample by means of a three-point ordinal scale. Whetherthese three categories of onset or the more usual dichotomousapproach was used, much predictive sensitivity was likely lost.Finally, despite its prominent status as a predictor ofoutcome, evaluations of the type of onset have not controlled forother well—known prognostic indicators. One would expect, forexample, a strong degree of intercorrelation between type ofonset, premorbid social functioning, and outcome (Cancro &Sugerman, 1968; Stoffelmayr et al., 1983). Similarly, the typeof onset may gain its predictive power because it signals a moresevere form of illness. In general, any assessment of the uniqueinfluence of onset on outcome would need to control for theinfluence of “third variables” such as age at onset, sex, andpremorbid social or occupational functioning.To summarize, research must incorporate a number ofmethodological requirements in order to achieve a clearassessment of the influence of onset on the course of illness.* To date, I have not been able to discover the origin of thetradition of using 6 months as the crucial point for definingacute vs. insidious onset.Literature Review 71These include: the use of a first-admission sample;distinguishing between two stages of onset, the prodromal phaseand the treatment lag—time; analyzing these intervals ascontinuous variables; and finally, controlling for other,overlapping predictors of outcome.Predictors of Short— vs. Longer—Term OutcomeThe schizophrenic process, even when disabling and chronic,does not involve continuous and unremitting deterioration.McGlashan’s (1988) review of long—term follow-up studies ofschizophrenia concluded that, at some point, the decline infunctioning appears to bottom out or reach a plateau. He wrotethat the bottoming-out point varies widely between individualsand across studies, but he estimated that it occurs on averageroughly 10 years after the onset of illness.In many domains, the best predictor of future behaviour ispast behaviour. This axiom is no less true vis-a-visschizophrenia where, for most measures of outcome, the strongestpredictive relationship is with the baseline value of the samemeasure of functioning. Of these baseline values, variablesrelated to premorbid functioning, socializing, occupationalfunctioning, and global ratings have been most frequentlyexamined. In general, the magnitude of the correlation betweenbaseline and outcome measures of functioning tends to be =.30-.40 (Stoffelmayr et al., 1983). Social functioning, however, hasa unique role as a predictor in that the magnitude of thecorrelation with outcome is remarkably consistent (L=.32—6)Literature Review 72across the domains of functioning that serve as measures ofoutcome. Previous work functioning, on the other hand, is morevariable, in that it correlates well with work functioning atfollow—up (average =.4O), but more modestly with other domainsof functioning at follow—up (average =.22 with hospitalization,and .24 with global assessment).McGlashan (1986) also offered evidence for the changing roleof predictor variables. He reported that, for the first decadefollowing diagnosis, variables relating to premorbid socialadjustment and intimate relationships were the best predictors ofoutcome. In the second decade, however, family environment andearly-illness characteristics attained predictive primacy. Inparticular, a history of overinvolvement between patient andfamily and symptoms such as paranoid ideation were the bestprognostic indicators. In the third decade and beyond, the bestpredictors of outcome were a family history of schizophrenia andthe premorbid acquisition of vocational skills. Unfortunately,the study’s retrospective design necessitated the use of asample that had at least two years of archival data available.This meant that McGlashan’s sample was biased towards chronicityat the outset, perhaps with a restricted range of outcome, andlimited generalizability. Nonetheless, it shows phase-relatedspecificity in predicting the outcome of schizophrenia.Even if variables do retain their predictive power atdifferent phases of the illness, one would not expect thatsimilar magnitudes of association will be present over differentpredictive intervals. Thus, it may be important for follow—upLiterature Review 73studies to identify the outcome period and exercise caution ingeneralizing the relevance of predictors to outcome “as a whole”(McGlashan, 1986).That there are different predictive relationships at variouspoints in the course of illness may have important implicationsfor intervention. For example, to demonstrate that the naturalhistory of the disorder includes a plateau may point to renewedtreatment efforts--long after patients, family and service-providers have been resigned to a fate that appears to be moredisabling than it would be otherwise. Similarly, changes in therelative strength of predictor variables at different points intime may have implications for the relative emphasis on work,family, or social life at various points in the patient’s“career”. Finally, to demonstrate similar predictiverelationships at different points in time may have the simplefunction of providing a replication of earlier results, and henceserve to make the earlier conclusions more robust.Alternatively, if a variable is associated with outcome at onepoint in time but not another, the possibility of phase—specificpredictive relationships is introduced.Current Vulnerability-Stress ModelsIn the foregoing pages, frequent mention has been made of‘diathesis-stess’ and ‘vulnerability—stress’ as if there were oneLiterature Review 74unitary concept. It is worthwhile to review briefly two of themore frequently cited models, to show that elaborations of thevulnerability-stress concept are anything but unitary.In 1977, Joseph Zubin and Bonnie Spring announced “A New Viewof Schizophrenia”, where a large number of people are endowedwith a degree of vulnerability that under suitable circumstancesmight be expressed in an episode of schizophrenic illness.Different concepts of diathesis had previously been proposed, butonly as a function of variable genetic contributions (Gottesman &Shields, 1982; Meehl, 1962). The Zubin and Spring model addedtwo new elements: First, they argued that there could benumerous contributions to a person’s vulnerability, over andabove genetic inheritance. Any number of acquired propensitiescould be due to traumas, specific diseases, perinatalcomplications, family experiences, adolescent peer interactions,and other life events that would affect homeostasis. In thisway, they proposed an integration of models of etiology derivedfrom sociological, developmental, genetic, neuropsychiatric, andother traditions. Particular concepts or variables were notproposed as elements of the model, only that causal influencescould come from other than the genetic domain.Second, Zubin and Spring argued that vulnerability was astable, enduring trait, whereas schizophrenia was a state thatwaxed and waned. A person with a high level of enduringvulnerability would have many ‘challenges’ in daily living thatwould be sufficient to prompt a psychotic episode, whereas onewith low enduring vulnerability would require nothing short of aLiterature Review 75catastrophic event to precipitate psychosis. The challenges thatdisrupt adaptation may be endogenous, e.g. maturational changeswithin the organism, inadequate nutrition, or pathologicalresponses to infection or stress. Alternatively, exogenouschallenges may be life events or disruptions in one’s socialnetwork.A more elaborate model is provided by Nuechterlein (1987;Nuechterlein & Dawson, 1984). Figure 1 illustrates his “overallheuristic schema of schizophrenic psychotic episodes” (1987, p.306). It shows four classes of variables that contribute topsychotic episodes: personal vulnerability factors, personalprotective factors, environmental protective factors, andenvironmental potentiators and stressors. On some occasions,they interact to determine transient intermediate states (e.g., aprodromal period), which are then subjected to coping strategies.If coping is not successful, a psychotic episode will ensue.The Nuechterlein (1987) model is an advance over that proposedby Zubin and Spring (1977) in that it is specified in greaterdetail. Each node in Figure 1 corresponds to one or moreempirically demonstrated vulnerability factors. Unfortunately,none of the risk indicators are pathognomonic for schizophrenia;thus, Nuechterlein points to a need for several separatevulnerability-stress models to address subgroups of schizophreniathat likely have distinct (but overlapping) etiologies. Arelated need is greater attention to the distinction betweeninitial and later episodes, so that residual effects from theLiterature Review 76Figure 1. Nuechterlein’s (1987) Vulnerability—Stress Frameworkfor Possible Factors in the Development of Schizophrenic EpisodesPersonal Vulneraoflity FactorsLiterature Review 77first will not be confused with vulnerability ‘markers’ (Zubin,Magaziner, & Steinhauer, 1983) for subsequent episodes.The vulnerability models of Zubin and Spring (1977) andNuechterlein (1987) are offered as conceptual frameworks. Assuch, they have three prominent limitations. The first is thatthe ‘black box’ of ‘Interaction’ shown in the Nuechterlein figureis present for conceptual and not empirical reasons; theprincipal emphasis to date has been on delineating the propertiesof the various components. As yet, the interactions among thosecomponents have received less attention, as have thebidirectional influences of the model’s components (as indicatedby the double—headed arrows). A second limitation is that thesignificance for a vulnerability model of some of the componentsis itself not completely clear, e.g. the hyper- vs. hyporesponsive electrodermal patterns of schizophrenics. Third, boththe Nuechterlein and Zubin and Spring models pertain to onset andrelapse; they are not intended as vulnerability models of courseand outcome.SummarY and Discussion of Literature ReviewFor a field that is conceptually nested in a vulnerablitystress paradigm, the apparent absence in the schizophrenialiterature of such models of course and outcome is surprising.Nonetheless, as has been described in the previous review,varying amounts of published data are available on the individualLiterature Review 78psychosocial and biological components that may be combined intoa diathesis—stress model.Psychosocial components of a diathesis—stress model. Theassociation between enhanced health status and supportive socialrelationships is well documented. Research workers must addressthe complexity of peoples’ involvement with others (Coyne &DeLongis, 1986; Lieberman, 1985). We need to know what it isabout social relationships, from what kind of people, and in whatkind of situations, that enhances health.Psychotic disorders, however, have not been the subject ofextensive scrutiny. Of two studies that have examined directlythe effect of supportive social relationships on the course ofschizophrenia, both reported a positive relationship withoutcome. Only one of those reports (Erickson et al., 1989),based on MAP Project data, assessed the various components ofsocial relationships as they pertained to course and outcome.That study awaits replication.A proper understanding of the impact of social relationshipson the outcome of schizophrenia in particular has some uniquerequirements. Among the most important are controlling for theeffects of onset, using a broad and representative sample offirst—episode schizophrenics, and controlling for the effects ofother influences on outcome. Assessment of social relationshipsshould include elements from the network, transactional, andperception-based approaches.Literature Review 79Other elements particular to schizophrenia also need to beaddressed. The role of social relationships must be comparedboth with normal controls and a psychiatric comparison group.Moreover, if it is true that McGlashan’s (1986) assertion istrue that predictive relationships vary over time, then the roleof social relationships in predicting outcome should be assessedwith at least two follow-up assessments.In the study of life events (LEs), the most common researchparadigm is to assess events that are proximal to a psychoticepisode. Results from recent studies indicate that increasednumbers of life events have an Immediate, adverse effect, but noinvestigation has yet assessed the cumulative effect of LE5 overa longer period of time.Such an assessment must occur in the context of a high degreeof statistical or methodological control. One must control forfactors that influence outcome, such as baseline functioning, thenature of onset, and the usual demographic characteristics. Arepresentative sample must be used to avoid selecting only acuteonset patients or some unknown portion of patients whose onset Iseasily described. Without a sample based on the full range ofonset characteristics, we cannot know that the existing findingsare not an artifact of research design.An assessment of the effects of LEs, as they are moderatedby social relationships, has not yet been reported. This may bedue to the recency of the evidence regarding the effect of LEs.An assessment of such a two—factor stress-process would be asignificant contribution to the literature.Literature Review 80Biological influences on the course of schizophrenia. Ofmany possible biological indices that may be related to thecourse of schizophrenia, two prominent candidates are lateralventricle (LV) size and smooth-pursuit eye movements (SPEM).Although the precise mechanism that may link these anomalies to adetrimental influence on course is unclear, linking prognosis andLV size or SPEM dysfunction may provide a valuable clue to theunderlying pathophysiology.Lateral ventricular enlargement is probably present in asubstantial minority of schizophrenic patients. It is alsopresent in mood disorders, but it lacks the significance forpremorbid functioning and prognosis seen in schizophrenia. Themodest degree of ventricular enlargement and the substantialoverlap with nonschizophrenic samples will limit the magnitude ofany association between LV size and outcome. Such an expectationis consistent with the literature reviewed above, including thereport from the MAP Project based on short—term outcome. Thecurrent project provides an opportunity for partial replication,based on longer-term outcome and a greater degree of control overextraneous influences.In contrast to LV size, smooth pursuit eye movements areunder partial genetic control, and dysfunction seems to berelatively specific to schizophrenia. Previous efforts toinvestigate the possible association between smooth-pursuit eyemovements and outcome indicated a modest, nonsignificantrelationship with lifetime hospitalizations and poorer short—termoutcome. The current project has an opportunity to examine thatLiterature Review 81association in greater detail, based on longer-term outcome.Once again, one would expect an association to be modest--sincemany neural mechanisms are involved in the control of SPEM.A vulnerablity—stress model of course and outcome. Avulnerability—stress model as it pertains to course ofschizophrenia has not been reported. Reports of the interplaybetween biological and social factors with respect to treatmentinterventions have appeared, but only to demonstrate in animmediate and experimental fashion that interpersonal stressors(family member rated as high on Expressed Emotion) have ademonstrable effect on a patient’s psychophysiological status(Sturgeon, Turpin, Kuipers, Berkowitz & Leff, 1984).To demonstrate a diathesis-stress model of course andoutcome, social factors must be assessed at the onset of thedisorder, in order to argue that they may have a causal role.Supportive social relationships and stressful life events areconceptually well-suited as components of a diathesis-stressmodel. Similarly, the choice of these measures is supportedempirically, even if the body of literature pertaining toschizophrenia is limited.Among a host of candidates, the biological factors reviewedabove are among the most robust findings with respect toschizophrenia. Lateral ventricular enlargement is present amongboth schizophrenic and affective patients, but it seems to havesignificance for premorbid and post-morbid functioning only amongschizophrenics due to its presence in the “Type II” syndrome ofCrow (1980). Smooth pursuit eye movement dysfunction, on theLiterature Review 82other hand, is not present among patients with mood disorders.Even though recent estimates of reduced prevalence seem toindicate that only a substantial minority of schizophrenicpatients have impaired SPEM, such a dysfunction may be associatedwith inferior course and outcome due to impairment in abilitiesrelated to nonvoluntary attention. To the degree that each ofsocial relationships, life events, VBR, and SPEM are related toprognosis, their concurrent and interactive influences may becombined in a vulnerability-stress model of course and outcome.83IV. METHODS: THE M.A.P. PROJECTThe arkers d Eredictors of Schizophrenia (MAP) Project atthe University of British Columbia was initiated in 1981 throughthe joint efforts of Drs. Morton Beiser and William lacono. Theprincipal interests of lacono were the psychophysiologicalmarkers of schizophrenia, whereas Beiser was primarily concernedwith the psychosocial predictors of the course of schizophrenia.Data collection began in 1982, and the last of the five-yearfollow-up interviews were held in the late summer of 1989. Themethodological attributes of this study include a representativeschizophrenic sample, two comparison groups, reliable diagnosticprocedures, incorporation of collateral sources of information,and a careful selection of measures.SamplesThe full MAP Project includes three samples: a group withfirst—episode schizophrenia, a comparison group with first—episode affective psychosis, and a matched comparison group ofnormal volunteers. Since the objectives of this dissertationinvolve within-group predictions, and issues regardingspecificity between diagnostic groups, the normal comparisongroup will not be discussed further. When recruited, the indexparticipant from each of the three groups was asked to nominate afamily member or close friend to provide collateral information.Members of the schizophrenic and affective psychosiscomparison group were recruited using the same broad set ofMethods 84criteria. Chosen to ensure an overinclusive bias at the level ofthe referring agencies, the criteria were:a) at least one clearly psychotic symptom (e.g. hallucination,delusion, thought disorder, marked change in behaviour), Qb) two or more of: marked reduction or loss ofinterest in usual activities; reduced initiative and drive;deterioration in performance; social withdrawal; persistentself—neglect;c) the ‘a’ or ‘b’ criterion occurred within three months of thereferral to the research project;d) age 15-54 years; ande) residence in the Greater Vancouver area for six months.Exclusion criteria included: receipt of a prescription forantipsychotic or antidepressant medication prior to contact withthe referring agency; evidence of organic disorders (e.g.epilepsy); psychosis due to alcohol or drugs; and mentalretardation.The use of the broad inclusion criteria had two advantages.The first relates to minimizing potential selection bias at thelevel of the referring agent: the liberal criteria ensured thatthe full range of schizophrenic patients were brought to theattention of project staff, rather than just the narrow andstereotypic type of chronic, impaired schizophrenics. A secondadvantage was the “instant” creation of a psychiatric comparisongroup.A further effort at minimizing selection bias involved the useof a community-wide referral system. Frequently, recruitment isdone entirely from in—patient populations--resulting in samplesthat tend to have disproportionate numbers of patients withMethods 85severe forms of the disorder. To avoid this bias towardsseverity, the MAP Project “cast the net” to a wide range oftreatment services. In addition to all in—patient facilities,every out-patient service and all psychiatrists in privatepractice in Greater Vancouver were regularly contacted. Referralsources also included all university and college counsellingcentres, immigration counselling centres, and a 1-in-S sample ofall general practice physicians. Short of knocking on doors in aresidential survey, we covered as comprehensively as possible allof the Greater Vancouver area.A total of 300 psychiatric patients were identified. Ofthese, 175 signed a consent form to participate. Ninety-four ofthe referrals (31.3%) refused to participate, and an additional31 (10.3%) either moved out of the area or disappeared from thetreatment system before they could be contacted. Analysescomparing the 175 participants with the 125 referrals lost to thestudy revealed no age or sex differences*.ProceduresData relevant to this thesis were collected at four points intime. Intake interviews were conducted within three months of* Due to ethical constraints, the effects of this “pre—inclusionattrition” are largely unknown. We have age and sex data onlybecause research assistants were able to observe thesecharacteristics in approaching the referred patients for consent.Ethical considerations prohibited the collection of any otherinformation. The effect of pre-inclusion attrition seems to bevirtually ignored by investigators engaged in predicting thecourse of schizophrenia: of dozens of studies in the area, nonehas addressed this problem.Methods 86the first treatment contact. Follow—up interviews were held nineand eighteen months later, as well as five years after intake tothe study. At intake, each subject was asked to nominate afamily member or close friend who would serve as a collateralinformant; these “significant others” (SOS) were interviewed onlyat intake and at 18—month follow-up. The time—line for thevarious interviews held with index participants and their SOS isillustrated in Table 7.Once potential subjects had been identified by a referringmental health agency, a member of the project staff confirmedthat the inclusion and exclusion criteria had been met. Thepotential participant was then approached to gain consent for thediagnostic and psychosocial interviews. In the case of acutelypsychotic patients, this occurred only when the attendingphysician (or service provider) indicated that the person wascapable of providing informed consent. To qualify for the “firstadmission” criterion, patients must have had their firsttreatment contact not earlier than three months prior to theirreferral to the study.At intake to the study, a psychiatrist or registeredpsychologist administered the Present State Examination (PSE, 9thEd.; Wing, Cooper, & Sartorius, 1974) to all patients. At thistime, Master’s—level interviewers administered questionnairesregarding various psychosocial variables (described below). Theinterviewers also obtained collateral information from the SO andreviewed clinic or hospital records.Methods 87Table 7. Summary of MeasuresIntake interviews* Demographics* Onset characteristics* Present State Examination (PSE; 9th Edition)* DSM—III Axis I diagnosis* Supportive social relationships* Life events in year prior to first treatment contact* Smooth pursuit eye movements* CT scan: brain lateral ventricle size* Collateral information from significant other (SO)* Review of clinical records* Global rating (DSM-III Axis V): highest level ofadaptive functioning in year prior to intakeNine-month follow-up* Axis I diagnosis* Life events since discharge from first hospitalization,or since intake interviewEighteen-month follow-up* Axis I diagnosis re 18—months & ‘Best Overall’ diagnosis* Life events since previous (9-month) interview* Collateral information from SO* Global rating (DSM—III Axis V): highest level ofadaptive functioning since 9-month interviewFive-year follow-up* Occupational functioning* Social activities (where full interview wasconducted; see text)* Type of residence* Medication status* Treatment status* Global rating of adaptive functioning (DSM-III Axis V)in the past monthMethods 88Upon completion of the intake diagnostic and psychosocialinterviews, the subjects were asked if they would participate inanother component of the project. They were offered the optionof returning to lacono’s laboratory where severalpsychophysiological variables, including smooth pursuit eyemovement function, would be assessed. A total of 134 (77%) ofthe 175 psychiatric participants volunteered for the second stageof the study. This second stage took about 90 minutes.Upon completion of the psychophysiological protocol,participants were asked if they wished to participate in a thirdstage, the CT scan. This session lasted an average of 40minutes. A total of 90 participants, or 51% of the originalpsychiatric group, consented to the CT scan.Nine months after their first treatment contact,participants were contacted to take part in a 45-minute follow-upinterview. The interview included an abbreviated DiagnosticInterview Schedule (DIS; Robins, Helzer, Croughan, & Ratcliff,1981), which was later used to determine diagnostic status at thetime of the nine—month follow-up interview. Stressful lifeevents were also assessed at this stage.At 18 months, participants were contacted to take part in asecond, longer follow-up interview. At this time, a psychiatristor registered psychologist repeated the PSE diagnostic interview.A lay interviewer obtained from the participant extensive dataregarding treatment, medication, work or school performance,living situation, and the extent of socializing during thefollow-up interval. Interviews with SOs also provided follow—upMethods 89data. The diagnostic data and information regarding outcomestatus (collected from the index participant, the SO, and areview of medical records) were used to determine diagnosticstatus and global ratings at 18-month follow—up. A total of 127(73%) of the psychiatric group were successfully interviewed at18 months.Five years after their inclusion in the study, in thesummers of 1987, 1988, and 1989, participants were againcontacted. On these occasions, due to scarce resources, the fullPSE diagnostic interview was not used. In its place, the trainedlay interviewers administered an abbreviated version of the DIS.Other than this change, the full battery of questions from theprevious follow-up interviews was used to document the variousaspects of adaptive functioning and treatment utilization thattogether comprise outcome. The measure of social support,previously administered at intake, was also repeated at the five-year follow—up. The SO was not contacted at this time.Attrition is a problem for all longitudinal studies. In theMAP Project, two factors made the potential for attrition evenmore pronounced. First, because the sample was young (mean age25 years, SD=7 years) and predominantly male (68%), we expected alot of geographical mobility over the course of the study. Asecond factor interfering with the response rate at follow-up wasthat, during the early stages of the project, British Columbiawas gripped by an economic recession: the unemployment rate forthe young male population hovered near 25% (Shaw, 1985). At thisMethods 90time, many young people had to travel great distances to findwork.In light of the usual influences on attrition, as well as thetwo that were especially relevant to the MAP Project, projectstaff mailed out newsletters and Christmas cards over the courseof the follow-up period, and each time included change of addresscards. To further aid in relocating participants, both the indexsubject and the SO were asked, at intake and at 18 months, toprovide the names of two other persons who would likely knowtheir whereabouts, in case they lost contact with the project. Afinal effort at minimizing attrition was undertaken in thesummers of 1987 and 1988, when we attempted to locate bytelephone those participants who would be due for a five-yearfollow-up in subsequent years.In planning the five-year follow-up we thought that, due tothe greater passage of time, we might not achieve a sufficientresponse rate if we depended solely on face—to—face interviews.With this in mind, we developed alternate procedures that wouldserve as proxies, and could be used in several situations. Oneprocedure was for subjects who refused to participate in the fullinterview, but would agree to answer a few questions over thetelephone. Second, when we could not locate the subject, wecontacted the SO or another family member with whom we had hadprevious contact. Where the contact person had extensiveknowledge of the subject’s status, and where the subject wasunavailable, the contact person served as the informant.Finally, where we could not locate the subject, and where the SOMethods 91was not able to provide these few data, we relied on archivalsources, viz, clinical or hospital records. The data collectedin this proxy manner related to current status in four domains:treatment, medication, occupational functioning, and livingsituation. All of the “live” interviews for the five—yearfollow-up were completed by August 1989; the response rate isdescribed below.MeasuresThe measures, described below, are also summarized inTable 7. The descriptions of the instruments are divided intosections concerning those used at intake, at nine- and 18—monthfollow-up, and finally at five-year follow—up. Samples of thequestionnaire and interview measures that were used in the dataanalyses are included as Appendix I.Intake. The measures at intake pertain to demographiccharacteristics, baseline functioning, diagnoses, as well as tothe predictors of interest: social relationships, life events,eye movements, and brain ventricle size.* Demographic questionnaire. This instrument collected thestandard range of demographic information, includingoccupation, age, and sex. Since the psychiatric group wereyoung, and manywere still in high school or university,ratings of socioeconomic status (SES) were based on theirparents’ occupation. Using the Blishen socioeconomic indexBlishen & Carroll, 1978; Blishen & McRoberts, 1976), the SESrating reflects a combination of education, income, andMethods 92prestige. In all analyses, the SES variable for anyparticipant was taken as the higher of the two parentalratings.* Present State Examination (PSE, 9th Edition). The PSE(Wing et al., 1974) is a semi—structured diagnosticinterview, designed to yield reliable diagnoses. It wasadministered at intake and at 18—month follow-up by either apsychiatrist or a registered clinical psychologist.* DSM-III Diagnoses. Case conferences, based on the PSEinterview and collateral information, were held to determineDSM-III (American Psychiatric Association [APAI, 1980)diagnoses. At least two clinicians were present for caseconferences. Structured checklists for each diagnosticcategory were used in order to maximize adherence to the DSMLII Axis I criteria. Following completion of the checklists,diagnoses were generated on all five axes of the DSM—III foreach member of the psychiatric group. Table 8 shows thesample size for the various diagnostic groups for all Axis Icategories.The DSM—III Axis V rating represents the highest levelof adaptive functioning in the year prior to intake. Itassesses both social and occupational functioning; ratingswere based on data provided both by the index participant andthe SO.Assessments of the formal properties of the DSM—IIIaxes that were conducted for previous stages in the MAPProject are satisfactory. For the Axis I diagnoses, weMethods 93Table 8. Sample Size at Three Stages ofIntake, by DiagnosisStage of intakePsycho-Psycho- physiol- CTDiagnosis social ogical scan(n) (n) (n)Schizophrenic 72 60 45Bipolar 44 35 23Major 30 26 11DepressionSchizo- 9 8 4phreniformSchizo- 7 2 3affectiveParanoid 7 1 3Other& 6 2 1TOTAL SAMPLE 175 134 90Note. This table displays “best overall” diagnoses, whichcombine data from intake and two follow-up interviews. Atintake, two cases who received a functional psychosis werelater diagnosed as organic psychoses (viz, drug-induced, andhyperthyroidism).aAtypical, Brief Reactive, or organic psychosis).Methods 94measured agreement among pairs of project clinicians based onthe first 15 cases entered into the study. The Kappacoefficients ranged from .74 to .96. Inter—raterreliabilities for Axis V1 the highest level of adaptivefunctioning in the year prior to intake, were .89 (based onPearson correlations).* Interview Schedule for Social Interaction (1851). Thisquestionnaire assesses both the structural characteristics ofthe social network and the nature of and satisfaction withsupportive social transactions. It was designed to queryeight areas of social relationships as theorized by Weiss(1974), and was standardized on a neurotic and arepresentative community sample (Duncan—Jones, 1981, 1981b;Henderson, 1980; Henderson, Duncan—Jones, Byrne, & Scott,1980). Unfortunately, the results of the ISSI with the MAPProject psychotic and normal control samples did not supportthe use of those eight scales: Cronbach’s alpha coefficientswere far from satisfactory. As a result, we generated ourown scales from the 1551 items. In our revision, socialsupport as a transactional concept is measured with twoscales: the availability of either close and confidingrelationships, or acquaintances. Two other scales measurethe perception of support, i.e. the degree of satisfactionwith each of confidants and acquaintances. The scales areshown in Table 9 (headings ‘B’ through ‘E’): the alphacoefficients are quite satisfactory (.70 to .89 in oursample). The structural characteristics of the respondent’sMethods 95Table 9. Interview Schedule for Social Interaction: Scale ItemsA. Social Network Size: Kin and NonkinNumber of “family and close friends” named in response to thefollowing questions (divided into kin vs. nonkin). Persons namedin response to multiple items are counted only once:* Person with whom you feel most comfortable talking frankly* All persons with whom you are close, fond of, attached to* People you count on for emotional or material support* A single, lasting intimate relationship?* An important person with whom you’re no longer in contact?* Person close to you who died recently?* One who knows you very well as a person?* A person you can lean on?* Someone with whom you share important feelings?* A person who comforts you in his/her arms?* People who regularly depend on you for help,guidance or adviceB. Availability of Close and Confiding Relationships* No. of friends who could come to your home at any time(unannounced), and still be welcome?* No. of friends whose home you could visit at any time* No. of family and friends with whom you talk frankly* No. of friends or family who comfort you in their arms* No. of people outside your home who really appreciatewhat you do for them?* No. of people who tell you that you are good at thingsC. Adeuacv of Close and Confiding RelationshipsFor each of the items listed in ‘B’, respondants are asked“Is this about right for you, or would you like more or less ofthis kind of relationship?” (scored on 1-3 Likert-type scale).D. Acquaintances: AvailabilityIntroduction: “Acquaintances are people you know a little butwho are not close friends”. Items include:* No. of people who you know just a little, e.g. you may notknow their names but you greet each other when you pass by* No. of people not close to you from whom you easilyask small favours* No. of acquaintances for whom y. do small favours* No. of acquaintances from whom you could expect helpin times of trouble* No. of acquaintances who can expect practical or materialhelp from you in times of trouble* No. of people who, when you get upset, you can tell themjust how you feelE. Adequacy of AcauaintancesRatings of adequacy of acquaintances, as per ‘C’.Methods 96social network are assessed with two indices, reflecting thenumber of kin and nonkin in the network.* Duration of onset. Three stages of the onset of psychosiswere determined: the appearance of first noticeable signs,the first appearance of florid psychotic symptoms, and theinitiation of treatment-seeking. The duration of theintervals between the first noticeable signs and floridpsychotic symptoms corresponds to what is often called theprodromal period. The interval between the first appearanceof florid symptoms and treatment—seeking we have calledtreatment lag—time. Derivation of the three stages of onsetis described in Beiser, Erickson, Fleming, and lacono (1991).Interrater reliabilities, based on judgements about thepatient’s age at each of the three stages of onset, rangefrom satisfactory to high. As described in Beiser et al.(1993), interrater agreement on the age at the onset of firstnoticeable signs, at first psychotic symptoms, and attreatment—seeking for the schizophrenic cases yielded Pearsoncorrelations of .66, .97 and .98, respectively. Coefficientsof agreement for the affective cases exceeded .90 for allthree stages.* Life Events. To assess the cumulative effects ofstressors, we used the Holmes and Rahe (1967) approach. Atintake, 9 months, and 18 months, subjects were interviewedwith a list of 33 possible events that comprised the SocialReadjustment Rating Scale (SRRS; Holmes & Rahe, 1967). Atintake, participants were asked to provide a briefMethods 97description of events that had occurred within the year priorto admission, and to indicate the month in which each eventhad occurred. The patient was also asked to provide arating of the impact of each event, on a 7—point Likert—typescale (—3=very negative; O=no effect; +3=very positive). Theoccurrence of life events was queried again at the 9- and 18-month interview, covering the period of time since theprevious interview.At each of intake, nine, and 18 months, the events wereclassified as to their temporal relationship with any majorpsychiatric disturbance. The classification schemedetermined if a reported event was independent or possiblyindependent of, or was likely caused by, pre—existingsymptomatology. As is the custom in the literature (Day etal., 1987; Ventura et al., 1989), only the ‘independent’ and‘possibly independent’ categories were subjected tostatistical analyses.* Smooth pursuit eye movement (SPEll) function. SPEM functionwas assessed using the procedures described by lacono andLykken (1981), wherein eye movements were recorded while theparticipant watched a luminous dot on an oscilloscope screen.The target moved in a sinusoidal fashion, covering twentydegrees of visual angle, at a frequency of 0.4 cycles persecond for 20 cycles. Head movement was minimized with theuse of chin, forehead, and temporal bone rests. Silversilver chloride electrodes applied at the outer canthi ofboth eyes recorded the position of the eyes compared to aMethods 98reference electrode attached to the right earlobe. Scoringof SPEM accuracy used the root-mean-square error method,whereby a computer program aligns the target and eyeposition, adjusts for phase and amplitude differences, andthen calculates the difference between the two signals. Theroot-mean—square error (RMSE) is then expressed in standarddeviation units.The RMSE approach has good formal properties. As aglobal index of SPEM function, it has good retest reliabilityover two years, shows evidence of heritability, anddiscriminates schizophrenic from other psychiatric andnonpsychiatric comparison groups (Szymanski, Kane, &Lieberman, 1991).* CT assessments. The CT scans were obtained in theRadiology Department of the Health Sciences Centre of theUniversity of British Columbia, using a Siemens Somatomscanner. A total of 13—16 slices were taken, each 8millimetres thick, at an angle parallel to the canthomeatalplane. Based on noncontrast scans, all measurements wereblind with respect to diagnostic status. The CT film thatshowed the lateral ventricles at their largest was chosen formeasurement, and enlarged to 6O of life size. The areas ofthe total brain and the ventricles were determined by tracingwith a planimeter the perimeter of the structures. The meanof repeated tracings was used for analyses. The inter-rateragreement on the VBR measures was high (intraclasscorrelation 0.92; Katsanis et al., 1991).Methods 99Nine—month follow-up measures. This follow—up interview isrelevant to this thesis to the extent that it was used to derive“cross-sectional” diagnoses on DSM-III Axis I, and to assess theoccurrence of life events.* The abbreviated Diagnostic Interview Schedule (DIS). TheDIS is a highly structured interview, designed to beadministered by lay interviewers (Robins, Helzer, Croughan, &Ratcliff, 1981). For the nine-month follow-up, only thesections relating to mania, depression, anxiety,schizophrenia, and paranoia were used.* Life events. The Holmes and Rahe (1967) Social ReadjustmentRating Scale was repeated at the nine-month follow-up, wherethe occurrence of life events since discharge from the indexhospitalization was queried.Eighteen—month follow-up. This interview replicated theearlier follow-up interview, with two exceptions. Rather thanthe DIS, a diagnostic interview was conducted by a projectclinician using the PSE. Supplementary data, regarding treatmentand medication status, occupational and social functioning, andresidential status, as well as a symptom checklist, were alsocollected from the index participants. At this time, SOsprovided collateral information. Once again, all of thesupplementary data helped to form diagnostic decisions.At this time, case conferences were again held to determinethe 9- and 18-month diagnoses. In addition to deriving these“cross—sectional” diagnoses, all of the data from the intake andMethods 100two short—term follow-up interviews were combined in caseconferences to produce a longitudinal, “best-estimate” Axis Idiagnosis. It is these “best overall” Axis I diagnoses, based onconsensus among project clinicians, that were used to definegroups in this project.Five-year follow-up. Five-year data were collected usingeither the full interview or abbreviated procedures. Asdescribed below, data from each procedure were combined toconstitute the dependent measure, the DSM-III Axis V (highestlevel of social and occupational functioning).* Full five-year interview. For about 50 of the originalsample, the full five—year interview was administered inperson. Similar to the nine-month follow-up assessment, itincluded the abbreviated DIS and the SCL-90 symptomchecklist. Occupational functioning was measured using theamount of time worked in the past month. Information aboutsocial functioning was collected by posing three questionsregarding socializing with each of friend and family.Specifically, each participant was asked to consider anaverage week in the past month and indicate: a) the numberof occasions, b) the number of people, and C) the amount oftime spent socializing in that week. Data regardingresidential status, involvement with treatment services, andmedication status over the previous year were also collected.These variables were also coded with respect to the monthprior to the interview.Methods 101* Abbreviated procedures. Where it was not possible toadminister the full interview, proxy procedures were used.As described earlier, one of the four substitute procedurewas comprised of a long-distance telephone call with theparticipant. Here, the data reflecting occupational andsocial functioning was the same as that derived from thefull, in-person interview. The other three proxy proceduresincluded: (a) a shorter telephone call with a participantwho refused the full interview; (b) an interview with aSignificant Other who had previously participated in thestudy; and (c) a review of clinical records. In thesubstitute procedures, the minimum data collected frompatients or significant others reflected four dimensions ofcurrent functioning (i.e., the past month):1. Are you working? (Please describe.]2. What is your living situation? [Please describe.]3. Are you seeing a doctor or anyone else foremotional problems? [Please describe.]4. Are you taking any medication? [Please describe.]* Five-year Axis V rating. Ratings regarding the highestlevel of adaptive functioning at five years reflectoccupational and social functioning in the past month.As a global measure, the Axis V ratings as described in 0511-III do not disaggregate qualitative from quantitativeaspects of either social and occupational functioning.Thus, there is an assumption that there will be at least amoderate degree of correspondance between quantitative andMethods 102qualitative components within each domain, and between thedomains themselves.Because the proxy procedures at five years did notalways gather data regarding social functioning, a smallmodification in the use of the Axis V scale was necessary.In the original 7-point scale, a person with moderateimpairment in occupational functioning would merit a ratingof either 4 (Fair) or 5 (Poor), depending on the concurrentsocial functioning. Similarly, a person with markedimpairment in occupational functioning would warrant arating of either 5 (Poor) or 6 (Very Poor), depending on thesocial functioning. For the cases whose proxy proceduresdid not provide data on social functioning (approximately45), intermediate ratings were used, e.g., 4.5 forparticipants with moderate impairment in occupationalfunctioning, and 5.5 for those with marked occupationalimpairment.The formal properties of the Axis V scale, based onintake and 18-month interviews, are good: earlier estimatesof interrater agreement resulted in Pearson correlations of0.89. Estimates of inter—rater reliability using themodified Axis V at 5 years are provided below, in theResults chapter.103V. HYPOTHESES AND ANALYSESThe aim of this dissertation project is to construct anintegrated model of the course of schizophrenia. Based on theMAP Project sample and research design, this project has thepotential to assess empirically four components of avulnerability—stress model as it applies to course and outcome.To the degree that the individual components are related to five-year outcome, and if the distributional properties of the dataare suitable, the components can be combined for simultaneousconsideration in predicting outcome.HypothesesThe specific hypotheses in this project follow from thegeneral aims described in the introductory chapter. Theyinclude:1. Social relationships will have a uniqueand independent role in predicting five-year outcome.2. Increased numbers of stressful life eventswill have a modest, negative effect on five-year outcome.3. Larger ventricles (ventricle to brain ratio; VBR) andsmooth pursuit eye movement (SPEll) dysfunction will beassociated with poorer five-year outcome for people withschizophrenia, but not affective psychosis.4. Life events and social relationships will concurrentlypredict five-year outcome, as will the interactionbetween the two. For social relationships, this isequivalent to hypothesizing a main and a buffering effectas they interact with life events.5. Social relationships, life events, VBR, andSPEll dysfunction will make concurrent contributions tothe prediction of five-year outcome.Hypotheses & Analyses104AnalysesBefore outlining specific analyses, some generalconsiderations are in order. In the prediction of five—yearoutcome, the previous literature review has elaborated the basesfor expecting relationships between and among the components ofthe proposed model: the control variables, the psychosocial andbiological predictor variables, and outcome. In general, I donot expect that the relationship between each of the independentmeasures and five-year outcome will be strong. Rather, I expectthat the control and predictor variables will have an importantbut modest relationship with outcome, since a vast array offactors influence the course of the disorder. Empirically, Iexpect that each of the modest relationships will be furtherattenuated by imperfect measurement, small variations inprocedure, within-group heterogeneity, and potentiallyconfounding factors that have not been assessed in this study.The general strategy is to use regression techniques toidentify unique and independent relationships between predictorvariables and outcome. The dependent measure is the DSM-III AxisV rating, the highest level of adaptive functioning in the monthprior to the five—year follow—up interview. The independentmeasures are comprised of a set of control variables (whichincludes the baseline Axis V rating, collected at intake to thestudy) and the set of predictor variables that represents thedomain of interest.Hypotheses c Analyses 105Table 10. Summary of Proposed AnalysesPhase 1. Methodological and Descriptive Analyses- response rate and differential attrition— choosing from multiple data sources at 5-year follow—up- formal properties of measures— treatment of missing values- sample characteristics: demographics, premorbidcharacteristics, and adaptive functioning at intake andfive years, for two groupsPhase 2. The Independent Role of Predictor VariablesThree separate hierarchical regression equations, each withthe 5-year Axis V rating as the dependent measure. For allequations, Step 1 is comprised of main effects for the predictorvariables of interest. Step 2 includes diagnosis and controlvariables previously identified as predicting outcome(L<.lO). Step 3 is comprised of the interactions betweendiagnosis and the predictor(s) of interest.Equation 1. Social relationships (N=100)Equation 2. Stressful life events (N=100)Equation 3. Eye movement dysfunction & brainventricle size (N=63)Phase 3. Data Reduction— social relationships (6 variables): principal componentsanalysis to reduce to one or two factors— life events (6 variables): sum the no. of independent &possibly independent events at intake, 9— & 18-months- control variables (age at onset, sex, SES, length of prodromalperiod, baseline Axis V): multiple regression of these,and diagnosis, to identify strongest predictorsPhase 4. A Stress-Process Model for Each Diagnostic GrouDOne hierarchical regression equation (N=100):Step 1. Main effects for social relationships, life events,and baseline functioningStep 2. Interactions between each social relationship andlife events variableStep 3. Main effects for diagnosis and baseline Axis VStep 4. Interactions between diagnosis and each of socialrelationship and life events variablesStep 5. Second-order interactions: socialrelationships by life events by diagnosisPhase 5. Concurrent Influences of Social and Biological FactorsOne hierarchical regression (N=63):Step 1. Main effects of social relationships, life events,VBR and SPEMStep 2. DiagnosisStep 3. Interactions: diagnosis with social relationships,life events, VBR and SPEMHypotheses £ Analyses 106The data analysis is based on a sequential strategy: asummary of the five phases is included as Table 10. In the firstphase, methodological and descriptive results are presented.Next, the hypotheses regarding the unique and independentcontributions of the each of the predictor variables areassessed. The third phase, if required, is devoted to datareduction. The fourth, if required, assesses a stress—processmodel within the psychosocial domain. Finally, if warranted, thepsychosocial and biological variables will be concurrentlyassessed in a combined vulnerability-stress model.Phase One: Methodological and Descriptive AnalysesThe methodological issues considered in this thesis pertain tothe aspects of the project that have not previously been examinedin the MAP Project. They include assessment of the formalproperties of measures, the response rate at five years, and aconsideration of the sources of data used in the five-yearfollow—up.In previous studies within the MAP Project, schizophrenic andaffective psychosis patients showed dramatic differences infunctioning, on a number of measures. Since both groups will becombined for some of the predictive analyses, descriptive datawill be presented to provide a sense of group differences inadaptive functioning at intake to the project and five yearslater. In this way, the different steps in the hierarchicalregressions that include diagnostic status will be morecomprehensible.Hypotheses c Analyses 107Phase Two: Establishing the Independent Role of PredictorsThis phase of analysis consists of several hierarchicalregression equations, where the first three hypotheses (statedbelow) are evaluated.As preliminary steps, descriptive data are presented for thetwo diagnostic groups for each predictor of interest.Demographic characteristics are also presented. In so doing, thereader will have a broad understanding of the manner in whichthese will affect multivariate predictions.The selection of control variables for entry in the regressionequation was determined by further preliminary analyses. Here,any control variable that was correlated with five-year outcomewas retained for the regression. A liberal alpha level (p<.10)was used to select control variables.For each of the hypothesized predictors, the five-year Axis Vrating serves as the dependent measure in a hierarchicalregression equation. In general, the regressions are based onboth diagnostic groups combined.Hypothesis One. Social relationships will have a uniqueand independent role in predicting five-year outcome.The first of the preliminary analyses pertaining to thishypothesis was an assessment of diagnostic differences in thesocial relationships at intake. Next, correlations provided anindication of the extent to which social networks, socialresources and the perception of social support are related tofive-year outcome, by diagnosis and for both groups combined. AHypotheses Analyses 108further preliminary analysis assessed the ability of the sixsocial relationship variables to predict outcome, withoutcontrolling for potentially confounding factors. In this way,the relative importance of social network, social resources, andperceived social support can be assessed, prior to the additionof control variables.In assessing the hypothesis that social relationships willhave a unique and independent role in predicting five-yearoutcome, the six social relationship variables were entered in ahierarchical regression as Step One. Diagnostic status,accompanied by control variables identified in preliminaryanalyses, comprised Step Two. The interaction between diagnosisand each of the six social relationship variables will compriseStep Three. If preliminary analyses suggest that any correlationbetween a control variable and outcome is substantially differentfor the two diagnostic groups, then a further interaction termmay be warranted.Hypothesis Two. Increased numbers of life events willhave a modest, negative effect on five-year outcome.The preliminary analyses again dealt with diagnosticdifferences in the predictor(s) of interest, in this case the sixvariables reflecting independent and possibly independent eventsat each of intake, nine, and 18 months. Preliminary correlationsbetween the six life event measures and five—year Axis V ratingswere computed for each diagnostic group and for both groupscombined.Hypotheses £ AnalysesIf significant predictive relationships are apparent in thesepreliminary analyses, then the life events variables can beentered into a regression equation where the DSM III Axis Vrating would again serve as the dependent measure. Step 1 wouldbe comprised of six life event terms. Step 2 would again adddiagnostic status and control variables identified in thepreliminary analyses. Step 3 would include six interactionterms: diagnosis by each of the life events variables.Hypothesis Three. Larger brain ventricles and SPElldysfunction will be associated with poorer outcome forpeople with schizophrenia, but not affective psychosis.The psychosocial factors were the subject of two separateequations, because of the large number of variables within eachdomain. Such is not the case in the biological sphere: each ofthe vulnerability factors is represented by one variable. Assuch, the unique and independent role of the two variables can beassessed simultaneously, in one regression equation.Preliminary bivariate analyses would assess any relationshipbetween outcome and each of the two biological vulnerabilityindicators. If the bivariate analyses indicate that VBR and SPEMpredict outcome, a hierarchical regression equation can be usedwhere VBR and SPEM measures comprise Step 1. Here, the SPEMvariable may be dichotomized into poor— and good-tracking groups.Step 2 would include diagnosis and control variables identifiedin preliminary analyses. The results from Step 3, comprised ofterms representing interactions with diagnosis, would determineif the diagnosis-specific aspect of the hypothesis is correct.Hypotheses & Analyses 110Phase Three: Data ReductionAlthough the sample size is reasonably large, there arenonetheless a limited number of variables that can be consideredin a multiple regression equation. If the analyses in Phase Twoindicate a larger number of significant predictor variables fromthe psychosocial and biological domains than can besimultaneously assessed in an hierarchical regression, then thefollowing data reduction techniques may be required.There are six measures of social relationships. Ifnecessary, they may be subjected to principal component analyses(PCA) to identify a small number of dimensions to use in (later)predictive equations. If the six variables show similarinterrelationships among the two diagnostic groups, then the oneor two components capturing the most variance would be retainedfor use in subsequent predictor equations. If the factorstructure for the two diagnostic groups is not similar, thenseparate PCA analyses can be conducted.In the life events domain, data reduction is a simpler task.If required, the six measures of life events can be summed toassess their joint relationship with five-year outcome.A third strategy may be necessary for control variables. Ifpreliminary bivariate analyses indicate that a large number ofcontrol variables are significantly associated with five-yearoutcome, then they may be entered into a separate regressionequation to identify those that make independent contributionsvs. those that are redundant. The control variables that achieveHypotheses & Analyses illstatus as independent predictors of five-year outcome may beretained for inclusion in Phase Three or Four.Phase Four: Assessing a Stress—Process ModelAnalyses in Phases Three and Four are contingent on resultsthat support the earlier hypotheses. In particular, the presenceof a significant relationship between life events and outcome iscrucial to both stress-process and diathesis—stress models.Hypothesis Four. Life events and social relationshipswill concurrently predict five-year outcome, as will theinteraction between the two. For social relationships,this is equivalent to hypothesizing a main and abuffering effect as they interact with life events.The test of this hypothesis requires three sets of predictorvariables: one representing social relationships, one for thecumulative LE rating, and one for the interaction term. Theexact number of variables in each set cannot be specified inadvance of the results from the Data Reduction phase.Previous regression analyses were based on all casescombined, and included first—order interaction terms involvingdiagnostic status (life events by diagnosis; social relationshipsby diagnosis). Since the assessment of a stress—process modelrequires terms representing the interaction between life eventsand social relationships, any diagnosis—specific tests requiressecond-order interaction terms (life events by socialrelationships by diagnosis).With these considerations in mind, the hypothesis concerning astress—process model would again use a hierarchical regressionHypotheses Lc Analyses 112strategy, with five-year Axis V as the dependent measure. Step 1would include social relationship and life events variables, andStep 2 would consider the (first-order) interaction between thepsychosocial elements. Step 3 would add diagnostic status andbaseline Axis V. Step 4 would be comprised of terms representingthe interaction between diagnosis and each of the socialrelationships and life events terms. Finally, Step 5 would becomprised of second—order interaction terms.Phase Five: Assessing Concurrently Social and Biological FactorsIt is not possible to anticipate the specific analyses atthis stage, since the components are largely influenced by thesignificance of earlier results. As such, what appears below isbut one example whereby the final hypothesis may be tested.Hypothesis Five. Social relationships, life events, VBR,and SPEW dysfunction will make concurrent contributions tothe prediction of five—year outcome.If results from Phase Two indicate significant relationshipsbetween each of the four predictor elements and outcome, thenStep 1 of an hierarchical regression might include five elementsrepresenting main effects: social relationships (two variables,from data reduction phase), life events, VBR, and SPEM. Step 2would include diagnosis, and Step 3 might include interactionsbetween diagnosis and the five terms from Step 1. Thesepredictors, together with the dependent measure, suggest 12variables in the Phase Three prediction equation——prior to theaddition of any control variables. Since the number of variablesHypotheses & Analyses 113is near the upper limit (63 cases: 12 variables), no controlvariables can be added in this analytic scenario.Formal Statistical ConsiderationsTwo prominent statistical considerations are in order. Thefirst relates to the inflation of the potential for Type I error,due to the presence of multiple parametric procedures, viz, fivemultiple regression equations. To control for Type I error inthe predictive analyses, the Bonferroni adjustment is used to setthe alpha level at .01.The second statistical consideration relates to power.Since the sample size is already set, I have determined theprobability of detecting an effect size of =.30. This modestmagnitude of effect is chosen because it is consistent with manyof the correlations described in the relevant literatures. Basedon one-tailed correlations and an alpha of .01, the power in theproposed analyses varies as a function of the number of availablecases. For equations involving social relationships and lifeevents variables, the number of cases will be about 100: here,power is .76. For equations involving eye-tracking function,where approximately 95 cases are present, power is .74. Finally,in the case of equations involving CT scan data (N=65), the poweris approximately .55. These calculations show that, for three ofthe four sets of predictors, the level of statistical power isadequate to detect correlations of .30. For the fourth, VBR,power is less than optimal.114VI. RESULTSBefore describing the predictive role of socialrelationships, life events, and biological vulnerability, someattention must be devoted to methodological issues andpreliminary, descriptive results. Phase One, immediately below,addresses several methodological issues. This section alsoreports descriptive results, including the demographiccharacteristics of the two diagnostic groups, as well as theiradaptive functioning at baseline and over the five-year course ofillness. Results in Phase Two consider the predictive role ofeach of social relationships, life events, and biologicalvulnerability indices. As will become apparent, the lack ofsupport for hypotheses regarding the independent role of some ofthe predictors precludes the need for progression to PhasesThree, Four, and Five, where stress—process and vulnerablitystress models would have been considered.To control for the inflation of Type I error, the Bonferronicorrection was applied to the analyses used to assess the fivehypotheses pertaining to the prediction of outcome. Here, fivetests are used; thus, the alpha level for statisticalsignificance is .01. In the predictive analyses, trends areidentified only if they meet the .05 alpha criterion. Whenconsidering preliminary and descriptive issues, no correction isapplied: the usual alpha level of .05 is used.Results 115Phase One: Methodological and Descriptive ResultsResponse RateData were considered eligible for the five year follow—up ifthey pertained to the period between 3.5 and 7.0 years afterintake to the study. This eligibility period was selected to bewithin two years of the fifth anniversary of the patients’ firstpsychotic episode; 3.5 years was used rather than 3.0 years so asto be nearer the fifth anniversary than the 18-month follow-up.Of the total MAP Project sample (schizophrenic, affective, andother psychoses), outcome is known for 133 of the surviving 168participants, for a response rate of 79.2%. Seven participantsdied during the five years following onset of illness: if oneconsiders death an analyzable outcome, the total response ratewas 80.0% (140/175). Of the 35 cases about whom data were notavailable, 11 (31%) refused to respond and 24 (69%) could not belocated.For the two diagnostic groups relevant to this project,there was no difference in response rate: outcome data areavailable on 54 (79.4%) and 55 (76.4%) of surviving schizophrenicand affective psychosis participants, respectively. There was nodifferential attrition based on age at onset of illness, durationof the prodromal period, or socioeconomic status. Finally, therewere no differences in baseline adaptive functioning, viz, theAxis V rating prior to intake. Overall, it appears that theschizophrenic and affective disorder patients for whom follow-updata were successfully collected can be taken as representativeof the entire (surviving) sample.Results 116As described in the preceding Methods chapter, four avenueswere available to assess five—year outcome status:a) the full interview;b) an abbreviated interview by telephone;c) collateral informants (i.e. a significant other); andd) medical records, which could reflect either psychiatric ornonpsychiatric care.Inherent in the use of archival sources is a a potential forbias towards severity: Poor-outcome cases present to clinics andin-patient wards more frequently than do good-outcome cases. Inthis study, however, there was an opportunity to minimize thisbias. When the only follow-up data available for this study camefrom such biased sources there was little choice but to use them,understanding their limitations. For this study, the ‘biased’designation is used when medical records described care given formental health purposes. Medical records reflectingnonpsychiatric care, e.g. emergency room visit for sprainedankle, are not considered biased.Frequently, however, data from multiple sources are available.In such cases (n=38, or 29), a set of rules is needed todetermine a priori which source to use for data analysis. Twoconsiderations are relevant in generating the rules: the date ofdata collection, and whether data came from biased or unbiasedsources. First, data are designated as ‘on due-date’ if theypertained to a period within three months of the fifthanniversary of intake to the study; if riot, they areResults 117Table 11. Use of Rules in Selecting for Analyses Data FromMultiple Sources Reflecting Five—Year Outcome.Rule1. On before offdue date.2. Unbiased beforebiased source.3. Same time period:combine.4. Use closest todue date.TOTAL 39 100.0 29.4% of cases % of allwith multiple cases withn 5-yr. data 5-yr. data15 38.5 11.310 25.6 7.55 12.8 3.89 23.1 6.8Results 118designated as ‘off due-date’. The second consideration is bias,as explained above. The following rules, then, form a hierarchy:1. Use data on due-date before off due-date;2. If all data are off due-date, use arbitrary (unbiased)before biased data;3. If all data are off due-date, and both arbitrary andbiased sources describe same time period, combine data;4. If all data are off due—date, and multiple data sets comefrom arbitrary or biased sources, choose data closestto due date.Table 11 shows the use of these rules in the 39 cases withmultiple sources of follow-up data. The use of these rulesresulted in 65.4 of follow-up data coming from the indexparticipant: 71 people completed the full interview, 14 theabbreviated interview, and two patients had their in—person datasupplemented by other sources. Psychiatric clinical records wereused in 216 of the cases (n=28), keeping potential bias to aminimum. A full description of the sources of five-year follow-up data used for analysis is shown in Table 12.The issue of potential vs. real bias as a function of thesource of follow—up data can be addressed by contrasting thosecases whose outcome status was determined exclusively through theuse of psychiatric records with those whose status was determinedfrom other sources of data. Here, the extent of bias cannot beknown fully; rather, we can only assess empirically indicatorsthat may reflect either bias or true differences between patientswho are more frequently in hospital. For example, if there areResul ts 119Table 12. Sources of Five—year Follow-up Data Used for Analyses.96 of surviving 96 ofSource of data n 5-year cases intake cases1. Full interview 71 53.4 40.6(index participant)2. Abbreviated interview 14 10.5 8.0(index participant)3. Significant other 7 5.3 4.04. Psychiatric records 28 21.1 16.05. Nonpsychiatric records 6 4.5 3.46. Multiple sourcesa re 7 5.3 4.0same time period7. Deceased 7 N.A. 4.08. No information 35 N.A. 20.0TOTAL 175 100.0 100.0a0f the seven cases with multiple sets of data describing five—year outcome, two included abbreviated data from the patient.Results 120archival/non-archival group differences in outcome, but none interms of baseline functioning, illness characteristics, orpremorbid status, then one could infer that bias indeed may bepresent. On the other hand, if differences in outcome wereparallel to pre—existing baseline differences, then one couldinfer that those receiving care were the more disturbed subgroup.In the combined schizophrenic and affective psychosis sample,the five-year outcome status of 21 cases (19.3%) is based onpsychiatric records alone. These patients have lower Axis Vratings at five years (mean s.d.: 3.3 + 1.3), compared to theremaining 88 “non—archival” cases (4.6 j 1.3; =4.41, df=107,p<.OO1). But important to the assessment of bias is that theyalso had lower baseline levels of adaptive functioning (meanintake Axis V ratings ÷ s.d.: 3.33 ±. 1.24 vs. 4.20 j. 1.17,=3.04, df=107, p<.OO5). The lower baseline functioning is notbecause the “archival” cases are more severely ill at intake:illness characteristics such as symptom severity, the length ofthe prodromal phase, and duration of the initial hospitalizationare similar for the two groups. Neither have the “archival”cases always been labouring at a disadvantage: educationalachievement and premorbid social functioning are strikinglysimilar for the archival and non-archival groups.A second way to assess bias is to examine the proportion ofparticipants whose outcome is assessed by means other thanpsychiatric records. Of these ‘nonarchival’ patients, 66% werereceiving psychiatric treatment at the time of the five—yearfollow-up. This suggests that, if the 21 ‘archival’ cases hadResults 121been ascertained by other means, 66% or 14 patients would havebeen in treatment. Thus, at most, the use of psychiatric recordsadded seven patients, or 6.4% of the combined schizophrenic andaffective sample, who otherwise would not have been located.Overall, these results indicate that a strong bias is not presentin the use of psychiatric records as an indicator of outcome.Formal Properties of MeasuresInter—rater reliability estimates for outcome measures werebased on 20 cases. Ratings were made, in the spring of 1992, bythe writer and a psychiatrist* at the Clarke Institute ofPsychiatry’s Division of Community, Culture and Health Studies.Since the global rating was derived using a 7-point ordinal-levelscale with intermediate ranks possible (e.g. 2.5, 4.5), anintraclass correlation coefficient (ICC) was used to estimatereliability. The result is an ICC of 0.95, indicating a highlevel of agreement between the raters.Reliability of the classification of stressful life events wasassessed by this writer and a doctoral candidate** in ClinicalPsychology at the University of Ottawa. Following a trainingperiod, 17 cases were randomly selected for inter—raterreliability estimates. Intraclass correlation coefficients werebased on the number of “independent” events per case, combiningevents across intake, nine—month, and 18-month interviews. The* Jiahui Zhang, M.B., M.Sc.** Dianne E. Chappell, N.A.Results 122ICC derived from this procedure was 0.73, indicating reasonablygood agreement between raters.In a previous paper (Erickson et al., 1989), we demonstratedacceptable levels of internal consistency for the four socialrelationships scales. There, Cronbach’s alpha coefficients werecalculated for the Availability and Adequacy of each of theAcquaintances and Confidants measures. At that time, we did notoffer evidence for the social network measures. Upon reflection,it seems that further assessment of the formal properties of thesix social relationships indicators (four scales, two tallies ofmembers of the network) is possible. Specifically, one canassess the convergent and discriminant validity by conducting aprincipal components factor analysis of the six indicators. Indoing so, we can begin with an inspection of the correlationmatrices for the two diagnostic groups, shown in Table 13.Table 13 shows that the magnitude of the inter-relationshipsamong the social relationship measures is moderate. There issome convergence, of the kind that one would expect: Thecorrelations between the two ‘adequacy’ measures for the twogroups are .44 and .52. There is less convergence than one wouldexpect in the way in which the number of nonkin is correlated toeither the availability of friends or the availability ofacquaintances among the schizophrenic group. Conversely, amongthe affective psychosis group the ‘nonkin’ variable is morestrongly correlated with the ‘availability of confidants’ scale.Overall, the moderate degree of convergence among the sixmeasures is adequate for factor analysis.Results 123Table 13. Intercorrelations Among Social Relationship Measuresfor Two Diagnostic Groups (schizophrenic patients below diagonal;affective psychosis patients above).No. of No. of Avail.of Adeq.of Avail.of Adeq.ofkin nonkin acquaint, acquaint. confid. confid.No. of .21 .13 .06 .28 .23KinNo. of .09 .13 —.10 .27 .00NonkinAvail, of .10 .11 —.25 .50 —.07Acquaint.Adeq. of —.01 .30 .19 .00 .44Acquaint.Avail.of .25 .53 .25 .33 .01Confid.Adeq. of .01 .25 .17 .52 .20Confid.Note. Sample sizes: schizophrenic and affective groups each N=66.Results 124Other indicators of suitability for factor analysis areevident in the correlation matrices shown in Table 15: There isreasonable similarity in the two matrices, there is a reasonableproportion of coefficients in the L=.25— 50 range, and there isat least one coefficient of a reasonable magnitude for eachvariable.Whether the diagnostic groups are combined or consideredseparately, other indicators suggest that these data areacceptable for factor analysis. The partial correlations aresmall, the Kaiser-Meyer-Olkin measures of sampling adequacy(Kaiser, 1974) for the two groups are in the 0.60—.70 range, andcommunality estimates are good. Using the principal componentsmethod of extraction, a two-factor solution is optimal (eitherfor all cases combined, or separately for the two groups). Thefactor loadings displayed in Table 14 are based on obliquerotation procedures. The results of the factor analysis, whetherbased on all cases combined or separately for the two diagnosticgroups, indicate convergent and discriminant validity for the sixsocial relationship indicators.The results of the principal component analyses indicate twoclear factors, whether derived from all cases combined or fromdiagnosis-specific analyses. In each case, the “meta—factors”are clear, and account for 59-60% of the variance. Although thefactors were not constrained to be orthogonal, there is littlecorrelation between the two factors in either patient group(schizophrenic sample =-.16; affective psychosis sample =.05).Table14.FactorAnalysesofSocialRelationshipsMeasures:CommunalityEstimatesandLoadingsfor‘Quantity’and‘Quality’FactorsforAllCasesCombinedandbyDiagnosis.AllcasesCOMMUN-LOADINGSALITYSchizophrenicpatientsCOMMUN-LOADINGSALITYQuant-QualityityAffectivepsychosispts.COMMUN-LOADINGSALITYQuant-QualityItyNo.ofkininnetworkNo.ofnonkinInnetworkAvail,ofconfidantsAvail,ofacquaint.Adeq.ofconfidantsAdeq.ofacquaint.%variance.72.83—.08.42 67•61.07—.13—.81.71.25—.77——39.2%19.6%.—.08.04.67—.08.85——34.7%24.8%Notes.Samplesizes:schizophrenicandaffectivegroupseachN=66.Factoranalysisbasedonprincipalcomponentsextractionandobliquerotation.U’Socialr’shipsQuant-Qual—indicatorityity.—.—.08.75—.——37.8%22.2%.72.85Results 126As indicated in bold type in Table 14, the Quantitydimension is comprised of four composite indicators: the numberof kin and nonkin in the social network, and the availability ofconfidants and acquaintances. The Quality dimension is comprisedof the two adequacy scales. The only perfidious variable is the‘number of kin in the social network’: When considered bydiagnosis, it loads on both the Quantity and Quality dimensionsfor each patient group.Within the Quality dimension, the number of family variablehas a diagnosis—specific characteristic. For schizophrenics, thenumber of family in the network is negatively associated withother quality indicators. In other words, schizophrenic peopleseem to be either satisfied with family and dissatisfied withpeople outside the family, or the reverse. Affective psychosispatients, on the other hand, report that they are generallysatisfied with both kin and nonkin or generally dissatisfied withboth.Overall, these principal component analyses are evidence forthe validity of the six social relationship measures used in thisstudy. They indicate the convergence and divergence ofquantitative and qualitative aspects of social relationships,supporting the conceptual distinction between social resourcesand social support. In particular, they offer support forvalidity of the two social network variables, whose formalproperties have not been previously assessed.Results 127Treatment of Missing ValuesIn studies such as this one, a small proportion of missingdata for each of a large number of variables can have aconsiderable effect on analyses. One way of minimizing thisdisadvantage is to use mean substitutions or similar procedures,provided only a small proportion of values are affected.In the first of the three broad categories of predictors,social relationships, data are missing for two reasons. Thefirst reason includes instances where a respondent failed toanswer a small number of questions on the ISSI, resulting inmissing values for one or two scales. Here, no more than 4.5?6 ofcases are missing on any of the six 1551 variables. It is thesevalues which were replaced by the respective mean for thediagnostic group. The second reason for missing data is thatparticipants (14 of 146, or 9.696) failed to complete the ISSIquestionnaire. Because of the relatively large number of cases,the ISSI data were left missing on these 14 cases.A similar approach was used in replacing missing values forthe stressful life events data. Missing data were not replacedif all of the three life events interviews (intake, nine months,18 months) were missed. For a number of cases, however, one ortwo sets of life events data were available: Here, the partialmissing data can be replaced using z—score equivalents. Forexample, if the intake life events total is missing, we cancalculate the z—score equivalent for the number of life events at9 and 18 months, and use the average z—score to calculate theintake value (based on mean and standard deviation of intakeResults 128values for the particular diagnostic group). This procedure wasused to replace missing data in 10 (6.9%) cases at intake, 8(5.696) at nine months, and 15 (10.4%) at 18 months.In the biological domain, one minor substitution procedure isindicated. Three of the participants who completed the CTassessment are missing eye—tracking data; for these cases themean RMS error value for the respective diagnostic group is used.The final considerations with respect to replacement ofmissing values refer to two control variables, the age at onsetof psychotic symptoms, and the duration of the prodromal period.For the first, 14 of 146 cases (9.6%) are missing the age atonset value: The replacement strategy here is to use theparticipant’s age at first treatment—seeking minus thediagnostically-appropriate median treatment lag-time. For thepatients whose date of onset we were able to determine, the lag—time is the period between onset of prominent psychotic symptomsand the first contact with a treatment facility. For the secondcontrol variable, the duration of the prodromal period, 14 cases(9.6%) are missing; the substitution uses the median for therelevant diagnostic group.Descriptive Results: Baseline and OutcomeTable 15 describes the groups in terms of their demographiccharacteristics and their baseline level of functioning. Resultsdescribing social relationships, stressful life events, andbiological vulnerability will be presented later, just prior tothe predictive results.Results 129Table 15. Sample Characteristics: Demographics and BaselineFunctioning by Diagnosis.AffectiveSchizophr. psychosisMean (S.D.) Mean (S.D.) ta probAge at intake (yrs.) 22.7 (5.3) 26.1 (8.0) 3.02 .005Age at onset of 21.6 (5.1) 25.9 (8.0) 3.84 .001psychotic symptomsDuration of prodromal 111 (141) 104 (177) .04 n.s.phaseb (weeks)Socioeconomic status 46.6 (16.1) 43.5 (13.7) 1.28 n.s.(Blishen scale)Baseline Axis V 3.5 (1.2) 4.5 (1.1) 5.54 .001(1=grossly impaired;7=superior)aAll t tests based on 144 d.f. bFor the ‘duration of prodrome’variable, the mean and standard deviation displayed are theactual values, although the test of group differences is based onVan der Waerden scores (normalized ranks).Results 130Considering all the baseline characteristics simultaneously,a preliminary multivariate analysis of variance (MANOVA)indicates that the two diagnostic groups are different(Hotelling’s T2=9.44, df=5,140, p<.O0l). Table 15 shows theresults from subsequent univariate tests. It shows that theaffective psychosis participants are slightly but significantlyolder than the schizophrenic patients at intake to the study andat the onset of psychotic symptoms. The proportion of men andwomen is also different in the two groups: men comprise 77.8€ ofthe schizophrenia group, compared to 60.86 of the affectivepsychosis group (Chi—square=4.93, df=1, p<.O5). There is nodifference in the level of education attained (data not shown) orthe socioeconomic background.The higher level of functioning in the affective psychosisgroup at intake is also apparent at five years. Figure 2suggests that there is a parallelism in the Axis V ratings of thetwo diagnostic groups over time. This observation is upheld by arepeated measures analysis of variance. With one between-subjects factor (diagnosis) and one within-subjects factor (twolevels of time), there is a main effect for diagnosis (F=44.1l,df=1,107 p<.OO1), a main effect for time (F=5.90, df=1,107,p=.O2), and no diagnosis by time interaction (F=.05, df=1,107,p=n.s.).A more clinically-oriented way to describe globalfunctioning is in terms of three categories. The category of‘Impaired’ includes persons with marked impairment in either workor social functioning (Axis V ratings of 1-3). The second> CD C) 1< CD -o tI) U) C) N -o-n (V -‘ (1)0DSM-IIIAxisV1’)()CD010)H 3 CDCD (a rtC 1 CDResults 132category, ‘Compromised’, describes patients who, despite amoderate degree of impairment in their social and occupationalfunctioning, are managing to continue with their activities (AxisV=3.5-4.5). Finally, the category of ‘Functioning’ describespeople whose Axis V ratings of the quanitity of either social oroccupational functioning is 5 or greater. The proportion ofschizophrenic and affective psychosis patients falling into thesecategories in the periods leading up to the intake and five-yearinterviews are shown in Table 16.Use of the three categories indicates that approximatelyone—fifth of people with schizophrenia are ‘Functioning’ ateither of the assessment periods. At the opposite end, betweenone—third and one-half are in the ‘Impaired’ category. Similarproportions are ‘Compromised’. Among affective patients almostthe reverse seems to be the case, with a small minority (15—26%)in the ‘Impaired’ category, and the majority (57-60%) in the‘Functioning’ group.Phase Two: The Independent Role of Predictor VariablesPrior to assessing the hypotheses, Table 17 illustrates thezero-order correlations between baseline characteristics andfive-year outcome, correlations that may overlap with thepredictor variables of interest.ResultsTable 16.. Proportions of Patients in Three Categories ofAdaptive Functioning Over Time, by Diagnosis.133SchizophreniaImpaired Compro- Funct—mised ioningAffective PsychosisImpaired Compro- Functmised ioningIntake () 48 32 20 16 27 5718 mos.(%) 63 20 16 26 28 465 yrs. () 32 46 22 15 25 60Note. Sample sizes at intake and five years: forschizophrenics, Ns=72 and 54 (respectively); for affectivepatients, Ns=74 and 55 (respectively).Results 134Table 17. Correlations Between Five-Year Outcome and DemographicCharacteristics and Baseline Functioning.Correlation with five-year Axis V ratingAffectiveSchizophrenics Psychosis All cases(n=54) (n=55) (n=109)**Age (yrs.): onset of .03 .16 .22psychotic symptomsSex (1=male;2=female) .11 —.04Duration of prodromal .13 .17k .l4phasea (weeks)** **Duration of treatment .08 -.34 —.23lag—time (weeks)Socioeconomic status .12 —.15 —.05(Blishen scale)Education (10—point .16 —.07 .08ordinal scale)Baseline Axis V .20k .21k 37**(1=grossly impaired;7=super ior)&For the ‘duration of prodrome’ variable, the correlation iscomputedbased oVan der Waerden scores (normalized ranks).p<.10 p<.O5 p<.OlResults 135Table 17 shows that characteristics observable at intake tothe proiect have at best a modest relationship to long-termoutcome. Indeed, within the schizophrenic group, two factors——age at onset and duration of the prodromal period--that haveachieved prominence as robust predictors of outcome inschizophrenia (e.g. Stoffelmayr et al., 1983) show littlerelationship to five-year outcome in this sample (L=.O3 and=.13, respectively). In the affective group, the onlycorrelation coefficient that clearly indicates a significantassociation with outcome is sex (L=—.32, p<.O1), where men arefunctioning better than women at five years*.In Table 17, two baseline characteristics predict outcome forall cases combined: age at onset and intake Axis V ratings. Forboth correlations, however, two concomitant effects are present.For each of the two variables there are: a) group differences atintake; and b) lower correlation coefficients with outcome wheneach group is considered separately. These observations renderthe higher “all cases combined” correlations a product of thedifferences between the two groups under consideration. Thisdoes not compromise validity; rather, it shows that diagnostic* When the affective psychosis group is subdivided by diagnosis,the men in the Bipolar group are functioning at a higher levelthan the women (H ± SD: 5.3 j. 1.0 vs. 4.2 16, respectively).In the Major Depressive group, however, the men and women arefunctioning at similar levels (H SD: men 5.3 ±. 1.2; women 5.01.2). Thus, the negative correlation between sex and 5-yearoutcome among the entire affective psychosis group is likely theresult of the lower functioning of the Bipolar women.Results 136status must be considered when analyses are based on all casescombined.Social RelationshipsIn this section, group differences are presented first. Theyare followed by simple predictive correlations, and then complexpredictive analyses.As described elsewhere (Erickson et al., 1989), substantialgroup differences in supportive social relationships wereapparent at intake to the study. Table 18 displays resultspertaining to social networks, resources, and perceived support,but first a reminder about scale construction is in order.As described previously, values for the each of the fourscales pertaining to social resources and perceived support werederived from the sum of standardized item scores.Standardization, based on the distribution from an age- and sex—matched normal comparison group, was necessary because the itemswithin each scale were expressed in different metrics. Thus, inTable 18 the ‘adequacy of close and confiding relationships’entry for schizophrenic patients indicates a mean of 1.4 standardunits below that of the normal controls.The data that appear in Table 18 for the two social networkvariables are not transformed: they are the actual number ofpeople in the networks of each patient group. For comparativepurposes, the social network of the normal comparison group wascomprised of (mean ± s.d) 6.3 + 4.6 nonfamily members, and 4.6 ±.2.9 family members (Erickson et al., 1989). Values for theResults 137Table 18. Social Networks, Social Resources, and PerceivedSupport at Intake for Two Groups.AffectiveSchizophrenic psychosissubjects subjects Fa prob(n=66) (n=66)mean s.d. mean s.d.Social networka) No. of kin 3.8 2.6 4.6 2.9 2.92 .09b) No. of nonkin 3.8 3.0 5.7 4.9 7.39 .01Social resourcesCa) Avail, of close —4.7 2.8 —1.4 4.2 27.48 .001& confiding rshipsb) Avail, of —3.2 2.4 —1.0 4.0 14.80 .001acquaintancesPerceiveda) Adeq. of close —1.4 3.5 —0.8 4.6 .59 ns& confiding rshipsb) Adeq. of —1.8 3.7 —0.3 3.5 5.81 .02acquaintancesaAll univariate F—tests based on 1,130 degrees of freedom.bsocial network values displayed in the table are the number ofpeople in the network. CData displayed are standardized scalevalues, based on a matched normal comparison group (see text).Results 138normal comparison group on the social resources and perceivedsupport scales are as follows (M s.d.): availability of closeand confiding relationships 0.0 j. 3.9; availability ofacquaintances 0.1 ± 4.4; adequacy of close and confidingrelationships 0.0 j 3.8; adequacy of acquaintances 0.1 ±. 3.4.Considering all six measures of social relationshipssimultaneously, significant differences were apparent between thetwo patient groups (Hotelling’s T2=7.19, df=6,125, p<.OO1).Subsequent univariate analyses of variance (ANOVA5) for threequantitative indicators, number of nonkin in the network,availability of confidants, and availability of acquaintances,indicate that the schizophrenic group had fewer supportive socialrelationships, compared to the affective psychosis group, priorto intake. A similar, though nonsignificant, trend was evidentfor the number of family members in the social network.On the qualitative side, schizophrenic patients are lesssatisfied with their more distal relationships. No significantdifference is apparent in the level of satisfaction withconfidants.Predicting outcome. The foregoing results are informativebut are not directly related to the hypothesis at hand, viz, thatsupportive social relationships will make a difference inoutcome. Table 19 displays correlations between the five-yearAxis V rating and each of the six social relationship measures.Table 19 indicates that the zero—order correlations betweensocial relationship ratings and five—year outcome are, at best,Results 139Table 19. Correlations Between Five-Year Outcome and Six Ratingsof Social Relationships, for Two Groups and All Cases Combined.AffectiveSchizophrenics Psychosis All cases(n=50) (n=50) (n=100)Social networkaNo. of kin—.19k .13 .O6No. of nonkin .31 .02 .21Social resource?a) Avail, of close .08 .20& confiding rshipsb) Avail, of .19 .08acquaintancesPerceived supportb*a) Adeq. of close .16 .10 .20& confiding rships*b) Adeq. of .11 .07 .17acquaintancesaSocial network values displayed in the table are the number ofpeople in the network. bata displayed under Social Resourcesand Perceived Support categories are standardized scale values,based on a matched normal comparison group (see text).* **p<.05 p<.OlResul ts 140modest in magnitude. A number of others are marginal inmagnitude, but nonetheless are in the expected direction.Specific significant findings indicate that, for all patientscombined, the availability of both confidants and acquaintancesis associated with five-year outcome. These correlations likelyreflect, in part, diagnostic differences in the two variables,since affective psychosis patients have both more socialrelationships and better outcome. Within the schizophreniagroup, there is a nonsignificant trend for a larger number ofnonfamily members of the social network to be associated withbetter outcome (=.31, p<.O5).The collective role of social relationships. What aboutthe Joint role of the various indicators of social relationships?There is, after all, overlap among the six measures. Forexample, the correlation between the two “quality” measures(Adequacy of Acquaintances, and Adequacy of Close and ConfidingRelationships) is L=.47. Similarly, the mean correlation amongthe four “quantity” measures (Number of Kin, Number of Nonkin,Availability of Acquaintances, Availability of Close andConfiding relationships) is =.35. Beyond the issue of overlap,it would be instructive to assess the magnitude of the collectiveassociation between outcome and the domain of supportive socialrelationships. The multiple regression equation in Table 20illustrates the results of such an assessment.The results in Table 20 show that, without considering anyother influences, social relationships collectively account forabout 86 of the variance in five—year outcome (Step 1. AdJustedResults 141Table 20. Hierarchical Regression: The Collective Role ofSocial Relationships in Predicting Five-Year Outcome.Step OneBetaStep TwoT Beta TStep ThreeBeta TNotes. Sample size: N=100 (50 in schizophrenic group, 50 inaffective group). Outcome is measured by Axis V ratings, thehighest level of social and occupational functioning in the pastmonth (higher scores indicate better outcome).aMult.R.37; Adj R2=.08; Chg R2=.13; chg240 (df693); Adj =.20; Chg =.12; chg_.1493 (df=7,92), p<..OOlCMult.R.55; Adj R2=.19; Chg R2=.04; (df=13,86), p=ri.s.Step 1. Main EffectsaNo. kin in network —.055 —.54 —.059 —.61 —.363 —1.08No. nonkin in network .070 .62 .065 .64 .716 1.48Avail, of confidants .094 .81 .106 .77 —.484 —.97Avail, of acquaint. .108 .79 .017 .13 .443 .74Adeq. of confidants .094 .81 .078 .71 .241 .67Adeq. of acquaint. .070 .62 .028 .26 —.097 —.28Step 2. Diagnosisb(Schiz=1; Aff=2) .394 3.86*** .501 1.48Step 3. InteractionscNo. kin X dx. .409 1.03No. nonkin X dx. —.736 -1.41Avail. confid. X dx. .561 1.20Avail, acquaint. X dx. —.452 —.80Adeq. confidants X dx. —. 191 -.54Adeq. acquaint. X dx. .106 .31P<.001Results 142R—square=.08, p=.03). No single social relationships measureaccounts for a significant amount of outcome variance. Afterentering the main effect for diagnosis in Step 2, the subsequentaddition in Step 3 of the interaction between socialrelationships and diagnosis does not add to the prediction ofoutcome.The near-significant collective contribution of socialrelationships may be due to the stronger contribution ofdiagnosis to the prediction of outcome. There is, after all,considerable shared variance between diagnostic status and socialrelationships, on the one hand, and diagnosis and outcome, on theother. The same argument may be made for other characteristicsobservable at intake to the study, i.e. the time when socialrelationships were assessed. The issue of whether socialrelationships have any unique and independent ability to predictoutcome--over and above diagnosis and other baselinecharacteristics—-is addressed below.The unique role of social relationships. Results from zero—order correlations presented to this point indicate that severalof the “control” variables predict five—year outcome, as doseveral of the social relationship measures. Moreover, thesecorrelations are often of differing magnitude for the twodiagnostic groups. Finally, the “control” variables and thesocial relationship measures are themselves inter—related.To assess the unique and independent role of these manyvariables, the following two—step multiple regression equationuses the five-year Axis V rating as the dependent variable. StepResultrs 1431 is comprised of main effects: the six measures of socialrelationships are accompanied by those “control” variables thatshow a significant or near—significant (p<.lO) zero—ordercorrelation with five—year Axis V ratings. Step 2 is comprisedof interactions between diagnosis and all six social relationshipratings, and interactions between diagnosis and controlvariables. However, prior to the addition of diagnosis/controlvariable interactions, there are 18 terms in the regressionequation. Thus, of all the potential interactions betweencontrol variables and diagnosis, only that between sex anddiagnosis has been added to the regression equation. Theinteraction with sex is selected because the difference betweenthe outcome-sex correlations for the two diagnostic groups isstrongest (schizophrenic group y.11; affective psychosis groupL=—.32; rff=.43) of all the zero-order correlations betweencontrol variables and outcome.Table 21 displays the effect of social relationships on five—year outcome, while controlling for concurrent influences fromage, sex, diagnosis, the duration of the prodromal period, andbaseline Axis V ratings. In Step 1, no social relationshipvariable achieves significance: Where five of six ISSI variableshad a significant zero-order correlation with five—year Axis V(Table 19), all have diminished to a nonsignificant independentrole. This appears to be due to their shared relationship withseveral control variables: Diagnostic status, duration of theprodromal phase, and baseline Axis V ratings each make aResults 144Table 21. Hierarchical Regression: The Unique and IndependentRole of Social Relationships in Predicting Five-Year Outcome.Step One Step TwoBeta T Beta TSTEP ONE: MAIN EFFECTSa.Social relationshipsNo. of kin in network —.067 —.69 —.637 —1.97No. nonkin in network .044 .43 .510 1.09Avail, of confidants .056 .40 —.174 —.35Avail, of acquaint. .004 .03 .326 .58Adeq. of confidants .034 .31 .164 .48Adeq. of acquaint. .054 .51 —.243 —.74Control variables*Diagnosis (1=sz,2=aff) .330 3.05 * .769 2.01Age at onset .070 .083 .79Dur’n of prodrome .238 2.48k .234 2.45Intake Axis V .242 2.10 .251 2.17Sex (1=male,2=female) —.152 —1.61 .459 1.45STEP TWO: INTERACTIONSb.Social rships X diagnosis*No. of kin in network .787 2.05No. nonkin in network —.528 —1.05Avail, of confidants .203 .43Avail, of acquaint. -.351 —.66Adeq. of confidants - .145 -.43Adeq. of acquaint. .301 .93Control variables by dx*Sex —1.014 —2.24Note. Sample comprised of 50 patients with schizophrenia and 50with affective psychosis.aMult.R.57; Adi R2=.25; Chg R2=.33; =394 (df=11,88), p<.OO1bMult.R.65; Adj =.30; Chg =.10; (df=18,81), p=.O8.* p<.05 ** p<.O1Results 145significant independent contribution to the prediction ofoutcome. This shared variance does not negate any associationbetween outcome and social relationships; rather, it indicatescovariance, in a manner where the ‘control’ predictors moreextensively account for variance in outcome that is alreadyexplained by social relationships predictors.Noteworthy among the significant main effects is thedirection of the sign for the “duration of prodromal phase” Betaweight. The positive weight (Beta=.243, T=2.53, P<.O1) indicatesthat, for all patients combined, a longer prodromal periodpredicts better outcome. When considered separately for theschizophrenic and affective psychosis groups, the direction ofthe correlations is positive but marginal (from Table 17: =.13,n.s. and y.17, p<..1O, respectively).Considering results regarding the interaction betweendiagnosis and predictor variables, two terms approachsignificance. The first, representing diagnosis by the number ofkin in the social network, indicates that there is a tendency forkin have a different effect on outcome for the two groups.Whereas the zero—order correlation between kin and outcome foreach of the two diagnostic groups is nonsignificant when taken onits own (Table 17), Table 21 indicates that the differencebetween the two correlations approaches significance (p<.05).A somewhat different interaction effect is seen for sex as apredictor. Table 17 showed that, among schizophrenics, sex isnot associated with outcome (=.11, p=n.s.). Contrarily, foraffective patients, sex is significantly associated with five—Results 146year outcome (=—.32, p<.O1; i.e. men have a better outcome).Step 2 of the current regression equation assesses the differencebetween =.11 and =-.32 and, after controlling for othervariables, finds that it approaches significance.In Table 21, the lack of significance for the nonkin—bydiagnosis interaction term is notable. The zero-ordercorrelation between the number of nonkin and outcome was r=.31(p<.05) for schizophrenics, but only L=.O2 (p=n.s.) foraffectives. By considering only these two correlations, onewould expect the interaction term in Table 21 to be significant.However, for all cases combined, the nonkin—outcome correlationwas =.21 (p<.O5); thus, the absence of an interaction in theregression equation may reflect a significant main effect, whichin turn may have been encompassed by the relationship betweenoutcome and baseline Axis V, diagnosis, or duration of prodromalperiod.To summarize, the results indicate that, for all patientscombined, the social relationships that are present at onset arerelated in a modest and nonsignificant way to five—year outcome,although this variance is shared with diagnosis and severalbaseline characteristics. Together, the social relationshipsvariables account for about of the variance in five-yearoutcome of both diagnostic groups combined. Within theschizophrenic group, there is a nonsignificant tendency for morenonfamily members in the social network to predict betteroutcome.Results 147Stressful Life EventsThe second hypothesis in this thesis was that increasednumbers of stressful life events will have a modest, negativeinfluence on five—year outcome. In presenting results related tothis hypothesis, the order will be familiar: first, descriptiveresults are presented regarding group differences in theoccurrence of life events in the time periods leading up tointake, 9-month, and 18-month interviews. Thereafter, simple andthen complex predictive relationships are presented in order toassess the role of life events on outcome.Table 22 displays the number of events reported by eachgroup in the periods leading up to the intake, 9—month and 18-month interviews. Tests of the group differences displayed inTable 22, as well as other parametric analyses of the life eventsvariables reported below, are based on 1og0(n+1) transformationsdue to the strong positive skew for each of the life eventsvariables. Nonetheless, for illustrative purposes, the mean andstandard deviations displayed in Table 22 are the actual numbersof life events.A preliminary MANOVA, with the number of independent and“possibly independent” events at the three points in time as thedependent measures, indicates that the schizophrenic groupexperienced fewer life events than the affective psychosis group(Hotelling’s T2=2..18, df=6,117, p<.05). The result isattributable to differences in the number of independent eventsat intake and 18 months. In the “possibly independent” categoryat intake, a nonsignificant trend is present (p=.O6), whereResults 148Table 22. Number of Life Events at Three Points in Time, byDiagnosis.AffectiveSchizophrenic psychosissubjects subjectsNumber of life events (n=64) (n=60) probmean s.d. mean s.d.IntakeIndependent 2.6 2.8 3.9 3.2 2.44 .02Possibly indep. 1.6 1.8 2.2 2.2 1.74 .069 month follow-upIndependent 2.3 2.5 2.5 2.6 .29Possibly indep. 1.5 1.8 1.5 1.8 .1418 month follow-upIndependent 2.8 2.5 3.6 2.6 1.84 .02Possibly indep. 1.5 1.6 1.6 1.8 .34 ——aThe tests, based on independent samples and pooled varianceestimates, use for analysis the 1og0(n+1) transformations of thenumber of life events. All tests have 122 d.f..Results 149schizophrenics experienced fewer events. The group differencesare not due to concomitant differences in baseline functioning orthe duration of the prodromal period: a multivariate analysis ofcovariance (MANCOVA) with these as covariates did not alter theresults described above.Predicting five-year outcome. Regardless of groupdifferences, does the number of events predict outcome? Table 23shows zero—order correlations between five—year Axis V ratingsand each of the life event indices.There is a complete absence of expected effects. Whereascorrelations of approximately =.3O had been hypothesized, theactual results, based on all cases combined, are remarkably closeto zero (range =—.O9 to =+.12). For schizophrenic patients,the correlations range from =-.O2 to =+.12. For this group,there is no single coefficient that departs significantly fromzero.The range of correlations among the affective psychosispatients is =-.O1 to =—. 31, indicating that life events do notpredict five-year outcome. The same conclusion is reached whenthe six life event variables are considered simultaneously in ahierarchical regression. As main effects, they do not predictoutcome (Mult. R=.l6, AdL R2=-.04, p=n.s). After addingdiagnosis in a second step (Mult R.=.46, Adi. R2=.15, p<.OO1),life events do not make a significant contribution whenconsidered in interaction with diagnosis (Mult. R.=.52, Adj.R2=.14, p=n.s.). When the correlations and regression resultsare considered together, it seems that there is no linearResults 150Table 23. Correlations Between Five-Year Outcome and the Log ofthe Number of Life Events, by Time and Diagnosis.Log of no. oflife eventsSchizophrenics(n=52)r prob.AffectivePsychos is(n=49)r prob.All cases(n=101)r prob.IntakeIndependentPossibly independent9—month follow-upIndependentPossibly independent18-month follow-upIndependentPossibly independent.04.08.09 ———.02 ———.02—.12—.16 ———.18 ——.12.02—.01—.09.05 ———.06 ——.10— .02—.01 ———.31 .01Results 151relationship between life events and five—year outcome for eithergroup.These results assess possible linear associations betweenlife events and long—term outcome. Nevertheless, it may be thata curvilinear association would be more appropriate, inparticular an inverted “U” relationship. If true, it wouldindicate that there may be an optimum amount of stressors,whereby either too few or too many events would be associatedwith a poorer outcome. To test such a notion, the bivariatescatterplots and the residual vs. predicted plots were examinedfor each of the log—life events by outcome combinations. Theresults indicated an absence of nonlinear relationships.Taken as a whole, the results derived here fall short ofthose hypothesized: the number of life events does not predictfive-year outcome for either group. Among schizophrenics, nozero—order correlation with five—year Axis V exceeds =.12.Among affective psychosis patients, only one life events variable(“possibly independent” at nine months) achieves significance,but it is not accompanied by an association between the“independent” life events variable at nine months and five—yearoutcome. Overall, it appears that, for both diagnostic groups,life events in the 2.5 years surrounding intake to the study donot predict five-year outcome.Biological Vulnerability and the Course of IllnessIf social relationships and life events comprise twocomponents of a diathesis—stress model of the course of illness,Results 152biological vulnerability indicators comprise the third andfourth. As described in the introductory chapter, previousstudies have failed to document correlations between outcome andventricle—to-brain ratio (VBR) or smooth pursuit eye movements(SPEM) among affective disorder patients. As such, thehypotheses here are that VBR and SPEM dysfunction predict outcomefor the schizophrenic group only.In presenting the descriptive results, a preliminary commentabout the distribution of the data is in order. Because eye-tracking error scores are left—censored (skew=+1.62), thisvariable was transformed using the log(n+1) convention. Thus,the group means and standard deviations for RMS error shown beloware actual scores, but the parametric analyses are based on log-transformed data. VBR data are displayed and analysed in theiroriginal, raw form. The results in Table 24 show that there areno group differences in either the mean eye—tracking error orventricle—to—brain ratio.As discussed in lacono, Moreau, Beiser, Fleming, and Lin(1992), there is a dramatic discontinuity in the distribution ofeye—tracking RMS error scores. Such a discontinuity may indicatea dichotomous phenomenon, and hence a dichotomous vulnerabilityfactor in predicting the course of a disorder. The distributionof the raw RMSE scores is illustrated in Figure 3 and shows thereare no cases between 304 and 353 units of error. Thus, thispoint of discontinuity is used to define good and poor eyetracking groups. When considered by diagnosis, there are 9 of 60(15.O’o) schizophrenic cases, and 5 of 61 (8.2%) affective casesResults 153Table 24. Eye-Tracking Errora and Ventricle—to-Brain Ratio byDiagnosis.Mean Std.Dev. (df) prob.Eye—tracking errorSchizophrenics (n=61) 185.38 121.63 (122) n.s.Affectives (n=63) 178.36 108.32Ventricle-to-brain ratioSchizophrenics (n=45) 6.60 2.62.98 (77) n.s.Affectives (n=34) 6.04 2.40aEye.tracking error using root-mean-squared-error approach.bThe test for eye—tracking RMS error is based on log(n+1)transformed data.Results 154Figure 3. Smoothed Distribution of Eye-Tracking RMSError Scores for 124 First—Episode Psychosis Patients.*No.ofpts.36 +33 ÷30 +27 4-24 +21 +18 +15 +12 +*9 ÷6 +3 +0 + + +*100 200 300 400 >480Eye—tracking root-mean—squared (RMS) errorResults 155with poor eye_tracking*. The difference in proportions is notsignificant (Chi—square=l.36, df=l, p=n.sJ.Unlike eye-tracking error scores, the VBR data are normallydistributed (skew=-.05; kurtosis=—.24). There is nodiscontinuity, only a smooth progression of data values. Thus,VBR is well suited to consideration in its continuous, raw form.The hypotheses are that VBR and eye-tracking error willpredict outcome only for schizophrenics. The results regardingVBR do not support the hypothesis: larger ventricles do notpresage poorer outcome in schizophrenia (L=.2O, p=n.s.).Similarly, the hypothesis of a significant correlation betweeneye—tracking error and five—year outcome for schizophrenics isnot supported (L=.ll, p=n.s.).Since the distribution of the smooth pursuit error scoressuggests a “natural” dichotomy, the outcome of normal and poor-tracking subgroups can be compared. Using a univariate , test,the results indicate that eye-tracking performance is notassociated with five—year outcome in schizophrenia. In fact,Axis V ratings are virtually identical for the good-trackinggroup (mean ±. sd: 3.8 ±. 1.3; n=42) and the poor-tracking group(3.9 £ 0.6; n=6; =—.29, df=46, p=n.s.). In sum, the resultsfail to support Hypothesis Three, that biological vulnerabilityfactors will predict long—term outcome.* These figures are slightly different than those appearing inlacono, !loreau, et al. (1992): In their paper, they useddiagnoses based only on intake data, compared to the current“best overall diagnoses” based on data from intake, 9-month, and18-month follow-up interviews.Results 156Building ModelsBoth the stress—process and diathesis—stress models of thecourse of illness require the hypothesized relationship betweenstressful life events and outcome. The results based on five—year Axis V ratings described above indicate a virtual lack ofcorrelation between outcome and the number of events. Hence,Phase Three, devoted to data reduction techniques, is notrequired. Similarly, Phases Four and Five, where the predictorswould have been combined into stress-process and vulnerability-stress models (respectively) are not required. Nonetheless, theresults pertaining to social relationships, life events, and thetwo biological vulnerability indicators can be discussed in thenext chapter with respect to their individual influence onoutcome.Summary of ResultsThe results regarding methodology suggest that proceduresinvolved in collecting the five-year follow-up data are sound.The response rate is good, with no detectable differentialattrition, and with minimal opportunity for bias due to thesource of follow-up data used in analyses. Preliminary analysesalso show substantial differences in the level of functioning ateach point in time for the diagnostic groups. Nonetheless, theResults 157chances in the level of social and occupational functioning overtime, as reflected by DSM—III Axis V ratings, are similar forschizophrenic and affective psychosis patients.In terms of the predictive results, none of the four sets ofpredictors is significantly associated with five—year outcome.Only one nonsignificant trend is apparent: Social relationshipscollectively account for about 8? of the variance in five-yearoutcome, although this variance is shared by several baselinecharacteristics (including diagnosis). When diagnostic groupsare considered separately, the hypothesized results are presentonly for schizophrenics, and only in the sense that the number ofnonkin shows a nonsignificant tendency to predict outcome.The hypothesis regarding the predictive role of life eventsis not supported. Group differences are present, where affectivepsychosis patients experience more independent events than doschizophrenics.Negative findings also apply to the biological vulnerabilityfactors. Contrary to the hypotheses, neither biological measurepredicts five—year outcome.The null results do not lend themselves to either a stress—process or a diathesis—stress model: These two models requirethat life events have a modest negative impact on course andoutcome. In this project, life events have no apparentrelationship with five—year Axis V ratings. As such, there seemsto be no point in trying to construct stress—process anddiathesis—stress models by integrating results that were foundResults 158here. Thus, the next step is to discuss in the next chapter theresults that have been presented——regardless of their place in anintegrated model.159VII. DISCUSSIONIn considering the foregoing results, the organization ofthe discussion will follow the same order as the previouschapter. The first section will review methodological andpreliminary results. The second will examine descriptiveresults, and predictive results vis—a—vis each of the threedomains. Finally, I will assess the significance of thisproject, its limitations, and discuss future research.MethodologySeveral methodological features may have affected thesubstantive findings. They include the definition of thegrouping variable, response rate and attrition, instrumentation,and formal statistical issues.DiagnosesIn the MAP Project, the diagnoses are based on thoroughprocedures. Information from semi—structured diagnosticinterviews, clinical records, and a collateral informant at bothintake and 18- month follow-up were combined at case conferencesto derive a “best overall” DSM—III diagnosis. The result isdiagnostic groupings that are as homogeneous as currentnosological standards allow.Discussion 160However pure, the DSM—III criteria provide for a narrow orrestrictive definition of schizophrenia. By defining only the‘nuclear’ or ‘core’ group of afflicted persons, the effect is tocircumscribe severity. Less restrictive criteria such as theICD—9 (World Health Organization [WHO], 1978) tend to includepatients with more affective symptoms, with psychotic symptomsthat are shorter in duration, or even with no psychotic symptomsat all (e.g. simple schizophrenia). Compared to DSM—III, onemight expect that the effect of less restrictive criteria wouldbe a broader range of outcome, particularly at the morefavourable end. A related result might be to increase themagnitude of any predictive correlations.This was the rationale behind an extensive set ofsupplementary analyses, using ICD-9 diagnoses to defineschizophrenic and manic-depressive groups. Use of the broaddefinition of schizophrenia resulted in a group comprised of 91cases; the manic-depressive group numbered 55. Nineteen of theDSM—III affective psychosis cases were designated schizophrenicin the lCD system. The results indicate that, descriptively, thechange in nosology has little effect on sample characteristicssuch as age at onset or baseline levels of adaptive functioning.Moreover, in assessing the predictive relationships between fiveyear outcome and each of social relationships, life events, VBRand SPEM, the correlations were substantially unaltered. Thus,the notion that the restrictive definition of DSM-IIIschizophrenia has restricted the range of adaptive functioning,and thus attenuated any prospective correlations that would haveDiscussion 161been present with a broader diagnostic definition ofschizophrenia, is unlikely.Using DSM—III criteria, the affective psychosis group isdiagnostically heterogeneous. As such, it is comprised ofapproximately equal numbers of patients with Bipolar or MajorDepressive Disorders, all of whom had psychotic symptoms in theindex episode. This heterogeneity is not an issue in this study,as the primary purpose of the affective patients was to act as apsychotic comparison group. In this thesis, analyses haveexamined the Bipolar and Major Depression patients separatelyonly when the pooled results appeared puzzling. This is the casein interpreting the role of sex as a predictor of five—yearoutcome for the combined affective group (cf. Table 17, andbelow). A detailed examination of the relationship between theoutcome of Major Depression and Bipolar Disorder per se withsocial relationships, life events, and biological vulnerabilityis beyond the scope of this thesis.ProcedureThe procedures in the intake and 18—month follow-up phases ofthe MAP Project have been reviewed in the preceding chapters.Several aspects of the five—year follow—up, however, are worthyof discussion here.The response rate was good: five-year outcome status isknown for approximately four-fifths of the patients. Thiscompares favourably to the earlier 9— and 18—month follow-uprates of 72-73. The five—year response rate also comparesDiscussion 162favourably with other naturalistic studies of first—episodecohorts (e.g. Strauss & Carpenter, 1977), although follow—upinvestigations that also provide treatment have higher responserates (e.g. Ventura et al., 1989).In this study, the epidemiological approach to recruitmentwas aimed at obtaining a representative sample. In the follow-upportions of the study, a high response rate is necessary at allstages to maintain that representativeness. With four-fifths ofpatients described at five years, and with numerous proceduresfailing to detect differential attrition, it seems that thesample described in this thesis can be taken as fairlyrepresentative of first-episode DSM-III schizophrenia.Although several procedures were undertaken to maintaincontact with the cohort, two improvements may have been possiblewith increased resources. Had we been able to expend more effortin the period between 18—month and five—year follow—up dates,fewer patients may have been lost to follow—up. Secondly, amongpatients whose outcome was determined, there may have been achange in the nature of the data used: Some of the patients whoprovided abbreviated data would have completed the full procedure(e.g., if funds were available for lengthy long-distance phonecalls).InstrumentationFormal properties of the various instruments were developedas part of earlier studies within the MAP Project, as well asspecifically for this thesis. Of the former, the internalDiscussion 163consistency coefficients for the social relationships variableshave the most bearing on this project. In Erickson et al.(1989), we described acceptable Cronbach’s alpha coefficients forfour of the ISSI scales. Although they are satisfactory, andconsistent with formal properties of other measures of supportivesocial transactions in the literature (Henderson, Duncan—Jones,Byrne, & Scott, 1980>, it may be that the true role of socialrelationships on the course of illness was attenuated due to lessthan perfect reliability, and thus is stronger than wasdemonstrated here.Indeed, the four ISSI scales were first developed for theearlier study assessing the role of social relationships onshort—term outcome (Erickson et al., 1989) because Henderson’soriginal ISSI scales (Henderson et al., 1980) did not cohere foreither our psychotic sample or the normal comparison group.Thus, the four new scales developed in the MAP Project need to bevalidated on another sample. It will also be important to reassess the evidence for convergent and discriminant validity ofthe six social relationship measures, using the principalcomponents analysis described earlier or other techniques.A similar point may be made for the formal properties of thelife events measures, where there is a satisfactory but notoutstanding level of agreement between raters. One way to raisethe inter—rater agreement about the life events data would be todate onset at the time of the original intake interviews, so thatlife event data collection could be anchored to onset. Forexample, any events that followed the onset of the prodrome wouldDiscussion 164automatically be deleted from consideration as independent of theillness.The dependent measure in the prediction equations, DSM—IIIAxis V ratings, has reasonably good formal properties. Theinter—rater reliability estimates at five years and at 18 monthsindicate a high level of agreement. Despite good reliabilities,the Axis V scale is a global rating. Even though it isrestricted to the domain of adaptive functioning, it nonethelesscombines two separate but related dimensions of outcome:occupational and social functioning. Ratings of any particularsample may achieve excellent interrater reliability, but thenature of a global rating is that correlations between it and adomain-specific predictor such as occupational or socialfunctioning alone will be attenuated compared to the use of acorresponding domain—specific measure of outcome (Stoffelmayr etal., 1983).Use of psychiatric records as the source of five—yearoutcome data seems to have had little effect on the results. Thepatients whose adaptive functioning was described by theserecords indeed had lower Axis V ratings than those ascertained byother means. Nonetheless, these patients also had lower levelsof functioning at intake. Thus, it seems that the use of apotentially biased source of data has not been translated intoactual bias.Statistical IssuesThe primary statistical consideration was controlling forDiscussion 165inflation of Type I error. The actual analyses depart somewhatfrom those proposed, but the number of statistical tests used toassess the hypotheses remained the same. The five analysesinclude:—- two hierarchical regressions where socialrelationships predict outcome (Tables 20 and 21);-- a regression equation where life events predictoutcome;—- a correlation between VBR and outcome; and-- a chi-square test assessing the effect of good andpoor eye—tracking on outcome.In light of the five statistical tests, the Bonferronicorrection dictated the use of an alpha level of .01.One of the four predictors of interest meets the conventional .05criterion. This one result, where the joint and uncontrolledeffects of social relationships account for 86 of the variance,is rendered nonsignificant with the use of the .01 criterion. Inshort, the use of a .01 alpha level reduces a somewhat mixed setof reliable results to a modest set of nonsignificant trends.Since the onset of the present study, I have added one stepin the analytic strategy. Earlier, the objective was to showthat a predictor had a unique and independent influence onoutcome. Here, the strategy was to show that socialrelationships or life events or biological risk added to theability to predict outcome, over and above that explained by anytraditional predictors such as age, sex, diagnosis, and level ofpremorbid functioning. Computationally, that meant that controlvariables would comprise the first step in hierarchicalDiscussion 166regression, and the predictor variable(s) of interest wouldcomprise a second step.Subsequently, I decided to examine more carefully thevariance that the hypothesized predictors (social relationships,life events, biological vulnerability) share with moretraditional predictors such as age at onset, sex, and premorbidstatus. This change in strategy is based on the notion thatshared variance is no less important as variance that is uniquelyattributed to a domain of interest. The result is an increasedemphasis on the zero—order correlations and, where warranted, apreliminary multiple regression equation where the jointinfluence of all indicators of the predictor domain of interestis considered (cf. Table 20). The change in strategy allowsgreater consideration of clinical implications. For example, ifsocial relationships or life events share variance with sex orage in causing variability in outcome, intervention can influencethe former but not the latter. This shift in analytic strategyincreases the ability to identify factors that are amenable tointervention.Descriitive and Preliminary ResultsAdaptive Functioning Over TimeApproximately one—fifth (16—22?6) of our schizophrenic sampleis healthy at any point in time, a proportion similar to thosedescribed in the literature (e.g., Ciompi, 1988). This fact willDiscussion 167be of some comfort to patients and their families, showing thatprognosis, even by DSM—III criteria, is not universally dismal.Another large portion of afflicted persons, approximately one-half, can look forward at the five—year mark to keeping theirdisorder at bay while they continue their daily activities.Members of this subgroup will need courage to meet the challenge,and help in their struggle. At the same time, many will beheartened by the fact that maintaining one’s activities andresponsibilities throughout the illness is not only possible, butprobable.To aid in the interpretation of the five-year results,supplementary analyses predicting 18-month outcome are presentedbelow. Figure 4 illustrates the Axis V ratings of the twodiagnostic groups at 18 months, as well as at intake and fiveyears.Both groups declined in functioning at 18 months, relativeto intake, and then increased to a level above both intake and 18months. Moreover, the parallel trajectory of adaptivefunctioning in the two patient groups over time, as seen earlierin Figure 2, continues to be apparent when 18—month status isincluded.The parallel course of the two diagnostic groups is notlimited to functioning measured by DSM-III Axis V ratings.Within the MAP Project, we (Beiser, Bean, Erickson, et al., 1993)have already shown that the parallel course phenomenon holds truefor occupational functioning between intake and 18 months. Inthat study, the amount of time devoted to work or schoolDiscussion 168Figure 4. Axis V Ratings at Three Points in Time, by Diagnosis.65F+-.134>Cl,Cl)020 -.—Intake Eighteen Months Five YearsTime31illAffective Pts. Schiz1Pts.Discussion 169was measured using a 95-point scale, where the major contributionto a patient’s score was determined by a tally of the number ofdays at a job or in school in the past 9-12 months. Thus, theparallelism was also detected with a domain—specific andcumulative measure, as well as a global ‘snapshot’ measure ofoutcome (Axis V ratings).Control Variables and OutcomeIt has been necessary to examine the relationship betweenoutcome and several demographic and baseline characteristics, soas to anticipate any “third variable” explanations for anyrelationships between outcome and predictor domains of interest.Two baseline characteristics of particular interest are theduration of the prodromal phase, and sex—related differences indiagnostic status.Sex differences. In the literature on course and outcome inschizophrenia, one of the most robust findings is that women dobetter than men (e.g., Goldstein, 1988). In the resultsdescribed above, however, this sex difference is absent: at fiveyears, the sex difference in ratings of current adaptivefunctioning is marginal and nonsignificant. Similarly, at 18months there is an absence of sex diflerences in outcome.Both findings, however, may be due to the index used tomeasure outcome. Axis V ratings are based on a ‘snapshot’definition of outcome. If a cumulative measure is used tomeasure short-term outcome, viz, occupational functioningDiscussion 170aggregated over the nine months leading up to the 18-monthfollow-up, the schizophrenic women in this sample show the usualadvantage over their male counterparts (Erickson, Beiser, Bean,et al., 1993). In the same domain, a ‘snapshot’ measure ofcurrent occupational functioning at five years (full-time vs.part-time vs. unemployed vs. in—hospital) indicates that womenand men show comparable outcome. Taken together, this mayindicate that periods of relapse and impairment are briefer forwomen, compared to men.Mode of illness onset. In the schizophrenia literature, theduration of the prodromal phase has been robust as a prognosticindicator (Stoffelmayr et al., 1983), where a lengthy, insidiousonset is a sign of poor prognosis. In our sample ofschizophrenic patients, however, duration of the prodrome showsno significant relationship with either five-year or 18—monthoutcome. The same is true for the affective cases. Whenconsidered in a multiple regression equation based on all casescombined (Table 21), however, the duration of the prodromebecomes significant (p<.Ol) in predicting five—year outcome.This indicates that the magnitude of the relationships detectedwith correlations (=.13 for schizophrenics; =.l7 for affectivepatients; all cases combined L=.l4) is significant when othervariables are assessed simultaneously. The surprising result isthe positive sign of the Beta coefficient, indicating that alonger prodrome predicts better five—year functioning.The positive relationship is difficult to understand. If itwas a null result, there could be several explanations for theDi5cussion 171discrepancy with extant literature. One relates to the nature ofthe samples described in the literature: With one exception,samples are comprised of multiple-episode patients. This meansthat poor—outcome patients will be over-represented, thusexaggerating any true association between mode of onset andoutcome. In the literature, the mean correlation between outcomeand mode of onset is a modest =.25 (s.d.=.08; cf. Table 6): Ifthis is exaggerated due to sampling bias, then the truerelationship may indeed be close to zero. A null relationship,however, is not what these data suggest: they point to a smallbut significant positive relationship.A related issue is that much of this literature has used theconcept of treatment lag-time as a proxy for mode of onset,instead of the duration of the prodromal period. If one useslag—time in our samples, we still get results that disagree withthe literature: Among the schizophrenic sample, correlationsbetween lag—time prior to first treatment contact and either 18-month or five—year outcome are virtually zero (L=.OB and L=.09,respectively).Many of the previous studies used a six-month cut—off pointto dichotomize either the prodromal period or the treatment lag—time, thereby defining acute- and insidious-onset subtypes (seeLiterature Review section above). When applied to our sample,this strategy provides consistent negative results. For example,dichotomizing the prodromal period, there is no difference infive-year outcome for the acute vs. insidious subgroups of eitherthe schizophrenic or affective patients (data not shown). ADiscussion 172similar absence of a significant relationship is found when thelag—time is dichotomized. Thus, the use of this less sensitiveanalytic strategy washes out the positive relationship betweenthe length of the prodromal period and five-year outcome.Predictive ResultsSocial RelationshipsWithin the domain of social relationships, the mostimportant new finding is a nonsignificant trend (p<.O5) toward adiagnosis—specific effect for the number of kin in the socialnetwork on five-year outcome. When all patients are combined foranalysis, the increased presence of family tends to have a morenegative influence on course for schizophrenic patients comparedto a more positive influence on affective patients. Althoughthis five—year multiple correlation is attenuated to ‘trend’magnitude, the result replicates the pattern shown at 18 months.For the schizophrenic patients, there is also a suggestionthat the number of nonfamily members of the social network atintake continues to have an influence on five-year outcome. Thisnonsignificant (L=.3l, p<.O5) relationship is similar to theinfluence shown at 18-month outcome (=.37, p<.O1), although thefive-year association is encompassed by other factors in amultiple regression paradigm.Discussion 173Overall, the magnitude of the association between socialrelationships and outcome has diminished between 18 months andfive years. Specifically, the correlations displayed in Table 19are attenuated compared to those reported in our paper describing18—month outcome (Erickson et al., 1989). A similar conclusionis derived from the multiple regressions predicting 18—month andfive—year outcome: Social relationship variables and theirinteraction with diagnosis accounted for 156 of the variance inshort—term outcome, but only 86 in longer—term outcome.In the literature, the only other study (Cohen & Sokolovsky,1978) assessing the effect of social relationships on outcomefound an effect among mildly but not severely impairedschizophrenic patients. Results from this sample suggest thatthe finding is not limited to mildly impaired patients.Together, the 18—month and five-year data are consistent with thenotion that social relationships do influence the course ofillness for schizophrenic patients with a broad range ofimpairment. Specifically, in light of the positive associationbetween nonkin and 18—month outcome for schizophrenic patients,and the positive association between kin and 18-month outcome foraffective psychosis patients, this may indicate thatschizophrenia implies a limited capacity to benefit from the kindof support that families members provide.Several other interpretations of these results are alsopossible. One is that family members of schizophrenic patientsmay be less able to provide constructive support in the face of adifficult burden. A third possibility is an interpretation basedDiscussion 174on an artifact: More relatives are associated with pooreroutcome for schizophrenia because more relatives are needed tocare for patients with more severe disorders. Thisinterpretation may be less likely than the others, since asimilar correlation was not apparent for the affective psychosispatients. In fact, the 18—month outcome/social relationshipsassociation was in the opposite direction for the affectivegroup.Life EventsIn the periods leading up to the intake and 18—monthinterviews, affective patients reported experiencing moreindependent events than did schizophrenic patients. Since thegroup difference remains after controlling for the length of theprodromal period and the level of baseline functioning, thisappears to indicate that it takes fewer stressors, on average, toprecipitate the onset of schizophrenia than affective psychosis.Alternatively, it may be that the nature of self—report issubject to differential bias in schizophrenia, compared toaffective psychosis. For example, the general sensitivity topeople and events after onset may affect the retrospectivereports of events prior to onset. Here, people withschizophrenia may count as an event something experienced astrivial by a person with affective disorder.It is likely that the lack of hypothesized associationbetween five-year outcome and the number of life events isDiscussion 175attributable to the 3.5 years between the latest assessment oflife events and the assessment of outcome itself. This lengthyspan of time may have reduced to zero what started as an effectof a modest magnitude. If true, it may be due to a truediminution in the effects of life events, or because a host ofother factors had opportunities to influence outcome. In eithercase, a stronger test of the role of stressful life events may bevis—a-vis short—term outcome.Life events and short-term outcome. Ratings of 18-monthoutcome differ from those of five—year outcome. The latterdescribes the highest level of adaptive functioning in the pastmonth, whereas the former selects and rates the best functioningover the past nine months. This raises an additionalconsideration regarding the temporal priority of life events andoutcome. If, for example, the best functioning during thefollow—up period was in month 10 or 11, then many life eventswill occur after the period reflected in the “18 month” Axis Vrating. Thus, only the intake and nine-month life events can beincluded in the prediction of 18-month outcome. Table 25displays the associations between 18-month outcome and lifeevents at intake and nine-month follow-up. Also shown arecorrelations between 18-month outcome and selected baselinecharacteristics.The table shows that, among schizophrenic participants, thenumber of independent life events occurring in the year prior tointake predicts 18-month Axis V ratings (=.38, p<.O1). WhileDiscuss ion 176Table 25. Correlations Between 18—Month Outcome’ and the Numberof Life Events or Baseline Characteristics, by DiagnosisSchizo—phrenics(n=55)r prob.AffectivePsychosis(n=54)r prob.All cases(n=109)r prob.NO. OF LIFE EVENTSIntakeIndependentPossibly indep.9—month follow-upIndependentPossibly indep..34 .01.14 .08BASELINE CHARACTERISTICSIntake Axis VSex (1=male,2=female)Age at onsetDur’n of prodrome3Treatment lag-time.37 .005.07 ——.12 ———.06 ——.09 ——.50 .001.28 .02.27 .02— .05 ——.22 .08.54 .001.24 .01.29 .001— .04 ——.00 ——Notes. Eighteen-month outcome is measured by Axis V ratings, thehighest level of social and occupational functioning in the ninemonths leading up to the 18—month follow-up (higher scoresindicate better outcome). Correlations are based on log10transformations. The duration of the prodromal period andtreatment lag-time are based on Van der Waerden scores(normalized ranks)..—.15.08 .16 .05—.01 ——Discussion 177the magnitude of the correlation is as predicted, the directionof the association is not: more events are associated withbetter outcome, rather than less. No other life events variablewas predictive of short—term outcome for schizophrenic patients.The correlations between life events at intake and 18-monthoutcome in the schizophrenic group may reflect a sharedrelationship with a third variable. The second portion of Table25 displays four of the leading candidates for ‘third variable’explanations: intake Axis V, sex, age at onset, and duration ofprodrome. Among the four, only the Intake Axis V predicts 18—month Axis V ratings (=.37, p<.OO5) for the schizophrenic group.Therefore, it remains possible that, for schizophrenics, thepositive correlation between the number of life events and 18—month outcome is an artifact of baseline functioning, i.e.previous impairment causes both fewer events at intake and poorer18-month outcome for the schizophrenic patients.A somewhat different picture is present among affectivepsychosis patients. As shown in Table 25, no single life eventrating is associated with 18—month outcome, although the‘independent’ classification at nine months shows anonsignificant (=.2O, p=.O8) trend. Again, the direction of thetrend is opposite to that hypothesized. No other correlationbetween short-term outcome and a life events variable approacheseven the “trend” magnitude.Among baseline characteristics, too, a different picture isevident in the affective psychosis group. Three of the fourDiscussion 178Table 26. Two Regression Equations: Stressful Life Events’ andBaseline Characteristics Predicting 18-Month Outcome2Separatelyfor Two Diagnostic Groups.SCHIZOPHRENIC GROUP (n=55) Beta T prob.Intake: # eventsIndependent .308 2.22 .03Poss. indep. -.079 —.61 -—9 months: # eventsIndependent —.001 —.01Poss. indep. .015 .10Baseline characteristicsIntake Axis V .285 2.12 .04Overall statistics for schizophrenic group:Mult. R.=.47; Adj R—Square=.15; F= 2.85 (d.f.=5,49), p=.O2.AFFECTIVE GROUP (n=54) Beta T prob.Intake: # of eventsIndependent -.043 -.35Poss. indep. .168 1.429 months: # of eventsIndependent .091 .78 ——Poss. indep. —.227 —1.85 .07Baseline characteristicsIntake Axis V .482 3.98 .001Sex (1=male, 2=female) .273 2.34 .02Age at onset .164 1.38 -—Overall statistics for affective group:Mult. R.=.65; Adi R—Sguare=.34; F= 4.89 (d.f.=7,46), p=.OO1.Notes. Number of life events is based on log transformations.Eighteen month outcome is measured by Axis V ratings, the highestlevel of social and occupational functioning in the 9 monthsprior to the 18-month follow-up (higher scores indicate betteroutcome). Intake Axis V is based on the highest level of socialand occupational functioning in the year prior to intake to thestudy.Discussion 179variables, Intake Axis V, sex, and age at onset, predict 18-monthoutcome. The finding that sex predicts outcome (i.e. women havebetter outcome) at 18 months is notable, given the oppositefinding at five years.In Table 26, two regression equations assess the unique andindependent role of life events as they predict 18-month outcomefor each diagnostic group. In each case, the dependent measureis the Axis V rating at 18 months, reflecting the highest levelof social and occupational functioning since the nine monthfollow-up. In each case, the predictor variables are the fourlife event variables and the baseline characteristics shownearlier to be correlated with short—term outcome.The most striking result in Table 26 is that concerning theschizophrenic group, where the predictive role of the number ofindependent life events at intake is diminished only slightly. Asbefore, the direction of the relationship is opposite to thatwhich was hypothesized: the sign of the Beta-weight indicates astrong tendency for a greater number of (log) life events topredict better outcome, over and above the baseline level offunctioning. Thus, the artifactual explanation mentioned above(e.g. that higher functioning participants get actively involvedin their world, welcoming the challenge of a greater number oflife events) is unlikely. Similarly, this relationship is not anartifact of a greater duration of illness prior to intake to thestudy, since Table 25 showed that the duration of the prodromalperiod was unrelated to 18-month outcome. As before, the lifeevents in the period leading up to the nine—month follow—up doDiscussion 180not predict 18-month outcome for schizophrenic participants.Together with Intake Axis V, the life events variables accountfor 15 of the variance in outcome (Mult R.=.47, Adjusted R—square=..15, p=.O2).Among affective psychosis participants, no life eventsvariable predicts outcome in the regression equation. WhereasTable 25 showed a nonsignificant trend toward a correlationbetween the nine—month “independent” rating and short-termoutcome (L=.2O, p=.08), there is no such result in the multipleregression analysis. The absence of either a nonsignificanttrend or a significant effect in Table 26 for the nine-month“independent” rating likely indicates that being female and/orhaving a higher level of functioning at intake encompasses the(nonsignificant) relationship between life events at nine monthsand 18-month outcome.In short, predicting 18-month outcome that is proximal to thelife events that were assessed in this study may be a morereasonable task than predicting distal, five-year outcome. Here,the number of life events does predict 18—month outcome, albeitdifferently for the two groups. For schizophrenics, life eventsin the year prior to intake do predict outcome: a greater numberof events is associated with better 18-month functioning, evenafter controlling for possible “third variable” explanations.Among affective psychosis patients, a nonsignificant trend isevident: The number of independent events at nine months show atendency to predict 18-month outcome (L=.2O, p=.O8), although thevariance in this relationship overlaps with the variance betweenDiscus5ion 181“control variables” and outcome. Again, the direction of thistrend is opposite to that hypothesized. Overall, it appears thateven this stronger, proximal test of the role of life events doesnot support a stress—process model of the course of illness foreither schizophrenia or affective psychosis.For schizophrenics, the meaning of a result opposite to thatwhich was hypothesized is unclear: one might speculate that(currently unknown) subtypes have varying thresholds for originalonset. Thus, high-threshold schizophrenic patients would requiremore stressors to precipitate illness onset, and subsequentlywould be less likely to relapse and more likely to have higheradaptive functioning throughout the course of the disorder.If this line of reasoning is valid, it would change themeaning of the terms in the regression equations in Table 26. Iftrue, the life events at intake would serve as an index of the‘threshold’, i.e. the individual baseline. Accordingly, thenine-month life events would be the only unequivocal index ofenvironmental stressors. Interpreted in this fashion, stressfullife events per se have no impact on 18 month-outcome.This interpretation points to the need for an analyticstrategy based on within-subject comparisons. Much of theliterature has used a repeated measures design, where the numberof events prior to a psychotic episode is compared with eventsfrom an earlier period. Such a method controls for individualdifferences with respect to stressor-distress thresholds, asnoted above.Discussion 182To summarize, two conclusions are possible from these short—term results. First, the relationship between life events priorto intake and outcome likely reflects an intra—individual abilityto withstand stress. Second, the months after a first psychoticepisode are a turbulent time. The lack of a correlation withoutcome likely indicates that events in those first nine monthsdo not affect the recovery process. To assess a stress-processmodel, it may be better to await later stages in the course ofillness.This study is not well aligned for direct comparison withother literature: No other studies of schizophrenic patientshave examined the effect of life events on the course of illnessin an aggregated fashion. One way of retaining the ‘aggregated’feature would be to assess life events as they relate to acumulative measure of relapse, e.g., the number of subsequenthospitalizations.Life events and onset. When the methods used to assess theeffects of life events on onset or relapse (as used in theliterature) are applied to these data, the results more or lesssupport the “triggering” hypothesis for stressful life events inthe onset of both schizophrenia and affective psychosis. To testthis hypothesis, the number of life events occurring within sixweeks of the onset of the prodromal period were contrasted withthe number of events in the preceding six weeks (i.e., 7—13 weeksprior to onset).Since the intake life events interview queried the year priorto the first treatment contact, only those patients whose onsetDiscussion 183occurred within that time period could be the subject of thisanalysis. The result of this constraint was a dramatic reductionin sample size (26 schizophrenic and 30 affective patients).After a further loss in sample size due to missing data, thefinal sample sizes were 18 schizophrenic and 21 affectivepatients.The mean number of life events in the two six-week periodsprior to the onset of the prodromal period are displayed inFigure 5. Inspection of the figure suggests differences in bothdiagnosis and over time.The distributions of the raw data shown in Figure 5, however,are skewed. Therefore, to assess diagnosis by time differencesin the number of life events, the raw data were transformed usingthe log(n+1) convention. Subsequently, a repeated measuresanalysis of variance (ANOVA), with one between-subjects factor(diagnosis) and one within-subjects factor (time), was conductedwith the log of the number of events as the dependent measure.The results indicate a main effect for diagnosis (F=10.76,df=1,37, p<.0O5) and a main effect for time (F=10.09, df=1,37,p<.OO5), but no interaction effect (F=1.13, df=1,37, n.s.).Subsequent paired i-tests indicate that affective patients hadsignificantly more (log) events in the six weeks leading up tothe onset of illness, compared to the previous six—week period(=2.94, df=20, p<.005). A similar test for the schizophrenicgroup indicates a nonsignificant trend towards a surplus ofevents in the period immediately prior to onset (=1.56, df=17,p=.O7). For the schizophrenia patients, the increase in theE > CD C-) CD -o C,.) C) N -oC,)CD CD (I) 0 -‘ 0 0) CD CD 7c- U) 0 -‘ 000(.11-01NumberofLifeEvents 1\)(A)010)01H 3 CDI-’.cn(QUi- (D I (D m (D U)Discussion 185(log) number of events in the weeks prior to onset, compared tothe previous period, represents an effect size of =.43; theeffect size for affective patients is =.65. Thus, it may bethat the failure to achieve significant results among theschizophrenic group is due to insufficient statistical power. Ifthat Is the case, these results suggest that the life eventshave a triggering role for stressors in the onset ofschizophrenia—-at least for those patients whose first treatmentcontact occurred within a year of the onset of the prodromalperiod. A similar conclusion may be drawn for the affectivepatients. Further, these results indicate that affectivepsychosis patients appear to be able to tolerate a higherbaseline level of stressors, compared to schizophrenic patients.Finally, there is some indication that schizophrenic patients donot require fewer (additional) stressors to precipitate psychosisthan do affective patients, since the interaction term in therepeated measures ANOVA is not significant.Lateral Ventricle SizeThe literature suggests that some unknown proportion (range7_949) of people with schizophrenia have enlarged ventricles.The best estimate for magnitude of enlargement is an effect sizeof .60, corresponding to 15—3O6 of original size.Early MAP results (Smith, 1986) show that neitherschizophrenic nor affective psychosis patients in this samplehave larger ventricles, compared both to a medical comparisongroup and a matched normal comparison group. Further, theDiscussion 186distribution of VBR in both patient samples is not bimodal,indicating an absence of a distinct subset with enlargedventricles. Finally, other MAP results have appeared recentlyindicating temporal stability in ventricle size: 15 of thefirst—episode schizophrenic patients were re-scanned after 1-3years, and no further enlargement was detected (Sponheim et al.,1991). In this light, these characteristics do not bode well forVBR as a predictor of outcome.As it turns out, VBR does not predict outcome at five years.No relationship is apparent whether statistical tests used VBR inits continuous form, or compared patients with the largest andsmallest ventricles (the strategy used by a number of studies inthe area).The absence of any association between larger ventricles andpoorer long—term outcome is surprising, since VBR does predictshort—term outcome in this sample. Katsanis et al. (1991)described the effect of large vs. small ventricles in outcome at9 and 18 months, controlling for baseline functioning. Theyconcluded that schizophrenics with larger ventricles had poorershort—term outcome*. If ventricular size predicts 18—month, butnot five-year outcome, then the biological vulnerability factorindexed by VBR may have its principal effect in the early stagesof illness.* If the median split approach is used to assess the relationshipbetween ventricle size and five-year outcome, no difference isseen between schizophrenic patients with larger vs. smallerventricles.Discussion 187In the schizophrenia literature, there are few establishedresults that unequivocally link brain structure with function.This is particularly true for the gross anatomy measured incomputed tomography: this technology is not capable of providinganswers to specific questions about etiology, and hence islargely unable to point to mechanisms by which the rather grossconcept of “biological vulnerability” used here might have aneffect on short—term, but not longer—term, outcome.Eve-Tracking PerformanceSmooth pursuit eye movement performance failed to show anassociation with five—year outcome. RMS error scores were notrelated to outcome when used in their continuous form ordichotomized at the “natural break” in the distribution.Nonetheless, eye—tracking does predict short-term outcome(lacono, 1988). Taken together, this suggests that the liabilityassociated with poor eye-tracking is more pronounced immediatelyafter onset.Using the empirically-derived cut—off described above, theprevalence of smooth pursuit dysfunction in this sample is muchless than previous reviewers suggested (5O—856 of schizophrenicpatients; Holzman, 1992). In this sample, depending on the dataused to determine DSM—III diagnosis, some 15—20% of theschizophrenics show smooth pursuit dysfunction, compared to 8% ofthe combined affective psychosis cases (11-12% of patients withDiscussion 188Major Depression*; 5—6% of those with Bipolar disorder). Sincethe use of these empirical techniques on several otherindependent samples has derived similar prevalence estimates(Clementz, Grove, lacono, & Sweeney, 1992), the true prevalencerate of SPEM dysfunction is likely much lower than that suggestedpreviously.Smooth pursuit eye—movement dysfunction was developed as agenetic marker, not as a predictor of course and outcome. Thedramatic reduction in prevalence in this and other recent samples(Clementz et al., 1992) will likely force a reassessment of therole of SPEM dysfunction in theories of etiology. With previousprevalence estimates of 50-85% in schizophrenic probands, and 45%in their clinically nonschizophrenic parents and siblings, therewas a strong prima facie case for a monogenetic contribution toetiology of all schizophrenic cases (Mathysse, Holzman, & Lange,1986). With quantitatively-derived prevalence estimates in theorder of 15—20%, SPEM dysfunction will retain its status as agenetic marker--but as a marker for a specific, familial subtypeof schizophrenia. Nonetheless, it will likely continue to beimportant to the etiology of this subtype.Biological Risk in GeneralWhile lateral ventricle size and oculomotor dysfunction arelikely independent, they are not the only two biological* Among the probands whose DSi1-III diagnosis is Major Depression(with psychotic features), smooth pursuit dysfunction is largelyconfined to the subgroup whose concurrent RDC diagnosis isSchizoaffective.Discussion 189anomalies in schizophrenia. The use of either SPEM dysfunctionor VBR makes no assumptions about other possible risk factorsthat may also influence the course of illness.That both biological factors examined here indicate aliability for short—term, but not longer-term outcome isintriguing. Since both VBR and SPEM performance are stable overtime, this may suggest the development of adaptation orcompensatory processes. It may reflect psychosocial adaptation,biomedical intervention, or more fundamental reorganization ofneurophysiology. In any case, the putative adaptation would taketime to develop, reducing the impact of any biological liabilityonly after the first year or two of illness. The biologicaldisadvantage may again become more prominent if adaptationalfactors diminish in later stages of illness.Models of Course and OutcomeAlthough conceptually compelling, merging the psychosocialand biological domains in schizophrenia may have better beenpostponed until the component parts were more firmly in place.The emphasis and rationale of this dissertation was primarily tointegrate the two domains, yet the role of life events—-althoughwell established in precipitating psychotic episodes--has not yetbeen well established as a predictor of outcome.Discussion 190Of the four predictive hypotheses, only that regarding therelationship between life events and outcome is crucial to thestress—process and diathesis—stress models. Since thatrelationship is not present in either the schizophrenic oraffective psychosis groups, assessing either model is notpossible.Significance of The ProjectThe original aim of this project, as proposed in June 1991,was to assess four sets of variables as they pertain to the five—year outcome of schizophrenia. To the degree that the fourhypotheses regarding the prediction of outcome were supported, afurther aim was to build an integrated, biopsychosocial model.Despite the general lack of support for the hypotheses, andsubsequent failure of such a model to materialize, valuableresults are present.Of the descriptive results, the parallel trajectory seen forthe two diagnostic groups strikes at a myth about schizophrenia:Clinical lore includes a profound pessimism about a decline inschizophrenia that is unmatched by other disorders. Theseresults show that, in comparison with psychotic affectivepatients, schizophrenic patients do not decline more; they simplystart off at a disadvantage, and decline or improve at the samerate in the early stages of illness.Discussion 191The parallel trajectories may also serve to highlight afundamental similarity between the two diagnostic groups: theirshared psychotic status. The early presence of psychosis maysignify a profound biological ‘shock’ that requires lengthyrealignment at the most fundamental levels: Post—psychosisneurochemistry may take time to re—establish homeostasis.Similarly, the psychosocial adaptation may be of comparablecomplexity, e.g. learning to draw on social support or developingnew coping styles. Finally, optimizing pharmacologicalintervention often takes experience over many months or severalepisodes.Of the hypotheses regarding the prediction of five—yearoutcome, one, pertaining to social relationships, receivedpartial support. The results indicate that friends and familypresent at the onset of illness continue to be important fiveyears later. While failing to achieve statistical significance,it is nonetheless a remarkable finding, given the lengthy passageof time. In particular, that specific components identified aspredictive of five-year outcome are consistent with thosepertaining to 18-month outcome provides a partial replication inan area that has seen little confirmation of results.That this association overlaps with other factors inpredicting five-year outcome is no less important than the uniqueand independent association of social relationships with shortterm outcome. Social relationships are amenable to interventionthroughout the course of illness, whereas sex, diagnosis, andbaseline functioning are not.Discussion 192The importance of the phases of illness is apparent in thedescriptive results. That there is first a decline in short-termadaptive functioning, and then a rise over and above baselinefunctioning, provides a longitudinal view of the early course ofillness. If nothing else, it provides support for the persistentefforts of afflicted persons and their families. It shows thatthe demands of dealing with the challenge of a complex disorderwill likely be greater in the earlier phases of the disorder.Moreover, the tremendous demands in the early phases will befollowed by better functioning in the later phases.In the biological realm, no previous study has examined theprognostic significance of VBR or SPEM at multiple follow-uppoints. That two putative indicators of biological liability,each stable over time, both predict short— but not longer-termoutcome, together point to possible compensatory processes asdescribed above.LimitationsWhile the MAP Project in general and this thesis in particularhave a number of strengths, there are also some notablelimitations. They include aspects pertaining to the researchdesign itself, the nature of some of the measures, and the sourceof data.Discussion 193The research design, though prospective and longitudinal, isnonetheless based on a “snapshot” approach to course and outcome.The dependent measure in this thesis, the DSM-III Axis V rating,and many measures of outcome in 9- and 18-month follow-upinterviews, reflect a cross-sectional view, and as such cannotcapture the rich and complex experience of illness that would besummarized in a “moving picture” view of course and outcome.Even as a snapshot measure, the Axis V--like all globalratings--combines the multidimensional concept of outcome into asingle dimension. Previous studies (Strauss & Carpenter. 1977;Stoffelmayr et al., 1983) have demonstrated that dimensions ofoutcome are best considered as “open-linked systems”. Forexample, it is possible for a person to have good functioning atwork but poor social functioning, or the reverse. Axis V, bycombining two dimensions into a single rating, loses complexityinherent in the lives of individual patients.As a global rating, Axis V rates primarily the quantitativeaspects of social and occupational functioning. For example,similar scores would be derived for a patient who had workedfull—time as a civil engineer prior to onset, yet who is onlyable to manage full—time work as a car wash attendant followingthe onset of schizophrenia. Occupational functioning is itself amultidimensional concept, where other aspects (not rated in AxisV) might include the complexity of the job and the quality ofperformance in fulfilling the task.Use of the Axis V rating in this particular data set indicatesone final limitation, in that five-year social functioning dataDiscussion 194were not available for approximately one-third of the cases. Theresult is that the dependent measure in the study relies moreheavily on occupational functioning than it otherwise would.Because the five-year data reflected only current functioning,we could not control for treatment and medication. Thelimitations of the abbreviated/proxy data at five years precludedeven a retrospective recounting of medication and treatment inthe past year (data that were collected in the full five-yearinterview).The final issue pertaining to research design is the amount oftime that passed between assessment of predictor variables andfive-year outcome. For example, any particular life event willlikely have its greatest effect within weeks of its occurrence.The results in this study likely indicate that the effect of lifeevents leading up to 18—month follow-up is negligible at fiveyears. Similarly, the effect of social relationships wasstronger at 18 months compared to five years, when the effect hasdiminished to a nonsignificant trend. Measurement of outcome atan intermediate point, e.g., three or four years, may have been amore reasonable effort. Alternatively, a reassessment of socialrelationships at an intermediate point would have beeninstructive.The nature of the life events measure used in this study, theSocial Readjustment Rating Scale (Holmes & Rahe, 1967), isanother limitation, inasmuch as it was designed for nonclinicalpopulations. Thus, there may be events of particularsignificance for psychiatric populations that are not measuredDiscussion 195with sufficient sensitivity; alternative instruments such as thePsychiatric Epidemiology Rating Interview (PERI) Life EventsScale (B.S. Dohrenwend, Krasnoff, Askenasy, & Dohrenwend, 1978)are available.The self—report nature of life events measures is anotherlimitation, in that they are susceptible to individualdifferences in the interpretation of what constitutes a minorlife event. Minor events such as a mild reprimand from asupervisor or a leaking waterpipe at home may or may not beinterpreted as a life event. At this end of the continuum, theoccurrence of an event is confounded with its meaning. In ourstudy, the use of trained interviewers and a structured formatminimized this danger.Future ResearchThere are a number of directions that one might take uponconclusion of this project, some of which are possible withinthis data set. In the psychosocial domain, it would beinstructive to examine the possible differential effects ofsocial relationships and life events on various subtypes inschizophrenia. Cohen and Sokolovsky (1978), for example, foundthat social relationships had no effect on the prognosis ofseverely impaired patients.Discussion 196If results from investigations such as these are to haveclinical applications, a more fine-grained analysis of socialrelationships data will be important. Here, we might considerthe structural characteristics of networks, e.g. networkmultiplexity or density as they relate to outcome. Given themagnitude of the prospective correlations with five-year outcome,this work would need to be based on 18-month outcome from the MAPProject, or on data from a new sample.The dual nature of the role of family members in the socialnetwork is intriguing. If the nonsignificant correlations withfive—year outcome are seen as supporting the significantcorrelations with 18-month outcome, there one can interpret thefindings as indicating only that greater numbers of kin show atendency to exert a small but significant negative influence onoutcome. Reversing the wording is probably more constructive forafflicted persons and their family: smaller numbers of kin arebeneficial, compared with larger numbers of kin. No evidence hasbeen presented that can comment on the effect of an increased ordecreased presence of any particular family member.Superficially, this result bears some similarity with the‘expressed emotion’ (EE) phenomenon (Leff & Vaughn, 1985), anavenue of investigation that must be explored in a futureproject. The similarity, however, is only superficial: the EEcomponents of ‘critical comments’, ‘overinvolved’, and ‘affectivestyle’ characterize the nature of a transaction, and do notcomment on the role of greater or fewer family members makingthose comments. A direct comparison of EE and the socialDiscussion 197relationships as measured here is necessary before claiming anyoverlap in the two concepts. Indeed, before measuring EE, apreliminary avenue for future research would be to assess thenetwork—related characteristics of any family members who do havea negative influence on outcome. Alternatively, it may be thatno individual family member has a negative effect; rather, it maysimply be that--for schizophrenic probands--family can sometimesbe too much of a good thing.If both social relationships and stressful life events hadpredicted outcome, a consideration of their interaction wouldhave constituted a stress-process model of the course of illness.Such a notion continues to be worthy of pursuit. A morereasonable assessment of the role of life events, however, wouldbe on short—term outcome. For example, life events could beassessed every six months over a two—year period. In this way,the interaction between life events and social relationships maybe assessed. In such a fashion, one could also consider personalresources such as coping styles, locus of control, or mastery asthey affect the relationship between stressors and distress overt i me.In the biological realm, a different strategy may be moreproductive. While brain morphology indicators have told us muchabout the likely primacy of brain dysfunction, and while smoothpursuit dysfunction may be instructive in understanding geneticetiology, neither phenomenon was developed to predict course andoutcome. Even if both predict short-term but not long-termoutcome, the results are difficult to interpret. It may be moreDiscussion 198fruitful to use investigative tools that have been designed torelate structure to function, e.g. neuropsychological assessmentor fast MRI.199REFERENCESAl Khani, M., Bebbington, P., Watson, J., & House, F. (1986).Life events and schizophrenia: A Saudi Arabian study. BritishJournal of Psychiatry, 148, 12-22.Alloway, R., & Bebbington, P. (1987). The buffer theory ofsocial support: A review of the literature. PsychologicalMedicine, II, 91—108.American Psychiatric Association (APA). (1980). Diagnostic andstatistical manual of mental disorders, (3rd ed.). Washington,DC: Author.Andreasen, N., Olsen, S., Dennert, J., & Smith, M. (1982).Ventricular enlargement in schizophrenia: Relationship topositive and negative symptoms. American Journal of Psychiatry,U., 297—302.Andreasen, N., Nasrallah, H., Dunn V., Olsen, S., Grove, W..,Erhardt, J., Coffman, A., & Crossett., J. (1986). Structuralabnormalities in the frontal system in schizophrenia: A magneticresonance imaging study. Archives of General Psychiatry, 43,136—44.Andreasen, N., Erhardt, J., Swayze, V., (1990). Magneticresonance imaging in schizophrenia: The pathophysiologicsignificance of structural abnormalities. Archives of GeneralPsychiatry, illAngermeyer, M., Kuehn, L., & Goldstein, J. (1989). Genderdifferences in schizophrenia: Rehospitalization and communitysurvival. Psychological Medicine, , 365—82.Bartfai, A., Levander, S., Nybeck, H., Bergrren, B-M., &Schalling, D. (1986). Smooth pursuit eye tracking,neuropsychological test performance, and computed tomography inschizophrenia. Psychiatry Research, j, 49—62.Beiser, M., Bean, G., Erickson, D., Zhang, J., lacono, W., &Rector, N. (1993). Biological and psychosocial predictors of iobperformance following a first episode of psychosis. Manuscriptsubmitted for publication.Beiser, M., Erickson, D., Fleming, J., & lacono, W. (1993).Establishing onset in psychosis. American Journal of Psychiatry,in press.Beiser, M., & lacono, W. (1990). An update on the epidemiology ofschizophrenia. Canadian Journal of Psychiatry, 35, 657-668.200Bern, D., & Funder, D. (1978). Predicting more of the people moreof the time: Assessing the personality of situations.Psychological Review, 85, 485—501.Benston, J., Reza, M., Winter, J., & Wilson, G. (1978). Steroidsand apparent cerebral atrophy on computer tomography scans.Journal of Computer Assisted Tomography, 2, 16—23.Besson, J., Corrigan, F., Cherryman, G., & Smith, N. (1987).Nuclear magnetic resonance brain imaging in chronicschizophrenia. British Journal of Psychiatry, jjQ, 161-163.Bird, J. (1982). Computerized tomography, atrophy and dementia:A review. Progress in Neurobiology, li 91—115.Bland, R. (1982). Predicting the outcome in schizophrenia.Canadian Journal of Psychiatry, 27, 52-62.Blishen, B., & Carroll, W. (1978). Sex differences in asocioeconomic index for occupations in Canada. Canadian Reviewof Sociology and Anthropology, , 352-371.Blishen, B., & McRoberts, H. (1976). A revised socioeconomicindex for occupations in Canada. Canadian Review of Sociologyand Anthropology, II, 71-79.Bleuler, N. (1978). The schizophrenic disorders: Long—termpatient and family studies (S. Clemens, Trans.). New Haven:Yale University Press.Bromet, B., Harrow, N., & Kasl, S. (1974). Premorbid functioningand outcome in schizophrenics and nonschizophrenics. Archives ofGeneral Psychiatry, Q, 203—207.Brown, G., & Birley, J. (1968). Crises and life changes and theonset of schizophrenia. Journal of Health and Social Behaviour,, 203—214.Brown, G., & Harris, T. (1978). The social origins ofdepression. London: Tavistock.Brown, G., Sklair, F., Harris, T., & Birley, J. (1973). Lifeevents and psychiatric disorders: Part 1. Some methodologicalissues. Psychological Medicine, 3, 74—87.Burstein, A., Adams, R., & Chapman, L. (1974). Prognosis inschizophrenia: A 5—year follow—up. Journal of Nervous andMental Disease, 137—140.Cancro, R., & Sugerman, A. (1968). Classification and outcome inprocess/reactive schizophrenia. Comprehensive Psychiatry, ,227—32.201Caplan, G. (1974). Support systems and community mental health.New York: Behavioural Publications.Carpenter, W., Heinrichs, D., & Wagman, A. (1988). Deficit andnon-deficit forms of schizophrenia: The concept. AmericanJournal of Psychiatry, 145, 578—583.Chung, R., Langeluddecke, P., & Tennant, C. (1986). Threateninglife events in the onset of schizophrenia and schizophreniformpsychosis. British Journal of Psychiatry, jj, 680-85.Ciompi, L. (1980). Three lectures on schizophrenia. BritishJournal of Psychiatry, JJj, 413—420.Ciompi, L. (1988). Learning from outcome studies: Toward acomprehensive biological-psychological understanding ofschizophrenia. Schizophrenia Research, 1, 373—384.Clementz, B., Grove, W., lacono W., & Sweeney, J. (1992).Smooth-pursuit eye movment dysfunction and liability forschizophrenia: Implications for genetic modelling. Journal ofAbnormal Psychology, 101, 117-29.Cobb, S. (1976). Social support as a moderation of life stress.Psychosomatic Medicine, , 300—314.Cohen, C., & Sokolovsky, J. (1978). Schizophrenia and socialnetworks: Ex-patients living in the inner city. SchizophreniaBulletin, 4., 546—60.Cohen, 3. (1977). Statistical power analysis for the behaviouralsciences. New York: Academic Press.Cohen, S., & Syme, S. (Eds.) (1985). Social support and health.Orlando, Florida: Academic Press.Coyne, 3., & DeLongis, A. (1986). Going beyond social support:The role of social relationships in adaptation. Journal ofConsulting and Clinical Psychology, j, 454-46.Crow, T. (1980). Molecular pathology of schizophrenia: Morethan one disease process. British Journal of Medicine, ZQ, 66-68.Crow, T. (1988). Aetiology of psychosis: The way ahead. In P.Bebbington & P. McGuffin (Eds.), Schizophrenia: The maiorissues, pp. 127-34. Oxford: Heinemann Publishers.Daniel, D., & Weinberger, D. (1991). Ex Multi Uno: A case forneurobiological homogeneity in schizophrenia. In C. Tamminga &S. Schulz (Eds.), Advances in Neuropsychiatry andPychopharmacology: Volume 1. Schizophrenia Research, pp. 227-35. New York: Raven.202Davidson, T., Bowden, L., & Tholen, T. (1979). Social support asa moderator in burn rehabilitation. Archives of Physical Medicineand Rehabilitation, Q, 556.Day, R., Nielsen, J., Korten, A., Ernberg, G., Dube, K., Gebhart,J., Jablensky, A., Leon, C., Marsella, A., Olatawura, M.,Sartorius, N., Stromgren, E., Takahashi, R., Wig, N., & Wynne, L.(1987). Stressful life events preceding the onset ofschizophrenia: A cross-national study from the World HealthOrganization. Culture. Medicine and Psychiatry, ji, 123-205.DeLisi, L., Goldin, L., Hamovit, J., Maxwell, E., Kurtz, D., &Gershon, E. (1986). A family study of the association ofincreased ventricular size with schizophrenia. Archives ofGeneral Psychiatry, il 148-53.DeLisi, L., Schwartz, C., Targum, S., Cannon-Spoor, E.,Weinberger, D., & Wyatt, R. (1983). Ventricular brainenlargement and outcome of acute schizophreniform disorder.Psychiatry Research, , 169-71.Dewan, M., Haldipur, C., Lane, E., Ispahani, A., Boucher, M., &MaJor, L. (1988). Bipolar affective disorder I. Comprehensivequantitative computed tomography. Acta Psychiatrica Scandinavica,77, 670—676.Docherty, J., Van Kammen, D., Sins, S., & Marder, S. (1978).Stages of onset of schizophrenic psychosis. American Journal ofPsychiatry, 420—426.Dohrenwend, B.P., & Egri, G. (1981). Recent stressful lifeevents and episodes of schizophrenia. Schizophrenia Bulletin, 7,12—23.Dohrenwend, B.S., Knasnoff, L., Askenasy, A., & Dohrenwend, B.P.(1978). Exemplification of a method for scaling life events:The PERI life events scale. Journal of Health and SocialBehaviour,,205—29.Dolan, R., Calloway, S., & Mann, A. (1985). Cerebral ventricularsize in depressed patients. Psychological Medicine, 15, 873—78.Duncan—Jones, P. (1981a). The structure of social relationships:Analysis of a survey instrument, Part 1. Social Psychiatry, ,55—61.Duncan—Jones, P. (1981b). The structure of social relationships:Analysis of a survey instrument, Part 2. Social Psychiatry, 16,143—149.Endler, N., & Edwards, J. (1988). Stress and vulnerability. InC. Last & M. Hersen (Eds.), Handbook of anxiety disorders, pp.278-92. New York: Pergamon.203Erickson, D.., Beiser, H., lacono, W., Fleming, J., & Lin, T.(1989). The role of social relationships in the course of first—episode schizophrenia and affective psychosis. American Journalof Psychiatry, )j, 1456—1461.Farmer, A., Jackson, R., McGuffin, P., & Story, P. (1987).Cerebral ventricular enlargement in chronic schizophrenia:Consistencies and contradictions. British Journal of Psychiatry,Q., 324—30.Finlay-Jones, R. (1988). Life events and psychiatric illness.In A. Henderson (Ed.), Handbook of Social Psychiatry, pp. 27-40.New York: Elsevier.Goldstein, J. (1988). Gender differences in the course ofschizophrenia. American Journal of Psychiatry, J.j., 684-89.Goldstein, J., Santangelo, S., Simpson, J., & Tsuang, M. (1990).The role of gender in identifying subtypes of schizophrenia.Schizophrenia Bulletin, 16, 263—75.Gottesman, I., & Shields, J. (1982). Schizophrenia: Theepigenetic puzzle. Cambridge: Cambridge tiniveristy Press.Greenblatt, M., Becerra, R., & Serafetinides, E. (1982). Socialnetworks and mental health : An overview. American Journal ofPsychiatry, U2, 977—84.Gureje, 0., & Adewumni, A. (1988). Life events and schizophreniain Nigeria: A controlled investigation. British Journal ofPsychiatry, 153, 367—375.Hammer, M. (1981). Social supports, social networks, andschizophrenia. Schizophrenia Bulletin, j, 45-57.Harding, C., Brooks, G., Ashikaga, T., Strauss, J., & Brier, A.(1987). The Vermont longitudinal study of persons with severemental illness: I. Long—term outcome of subjects whoretrospectively met DSM—III criteria for schizophrenia. AmericanJournal of Psychiatry, jjj, 727—735.Haug, J. (1962). Pneumoencephalographic studies in mentaldisease. Acta Psychiatrica Scandinavica, IQ (Suppi. 165), 9-106.Henderson, A. (1980). A development in social psychiatry: Thesystematic study of social bonds. Journal of Nervous and MentalDisorder, 63—69.Henderson, A. (1984). Interpreting the evidence on socialsupport. Social Psychiatry, , 49—52.Henderson, A. (1988). An introduction to Social Psychiatry. NewYork: Oxford linversity Press.204Henderson, A., & Brown, G. (1988). Social support: Thehypothesis and the evidence. In A. Henderson (Ed.), Handbook ofSocial Psychiatry, pp. 73-85. New York: Elsevier.Henderson, A., Duncan-Jones, P., Byrne, D., & Scott, R. (1980).Measuring social relationships: The Interview Schedule forSocial Interaction. Psychological Medicine, 10, 723—734.Hirsch, S., Bowen, 3., Emami, J., Cramer, P., Jolly, A., & Haw,C. (1993). The effect of life events and medication in theaetiology of schizophrenic relapse [Abstract]. SchizophreniaResearch, 9, 266.Holmes, T., & Rahe, R. (1967). The Social Readjustment RatingScale. Journal of Psychosomatic Research, 213-218.Holzman, P. (1992). Behavioural markers of schizophreniaresearch useful for genetic studies. Journal of PsychiatricResearch, , 427—45.Houston, 3., Maas 3., & Bowden C. (1986). Cerebrospinal fluidHVA, central brain atrophy, and clinical state in schizophrenia.Psychiatry Research, j, 207-14.Husted, J., Beiser, M., & lacono, W. (1992). Negative symptomsand the early course of schizophrenia. Psychiatry Research, jj,215—22.lacono, W. (1988). Eye movements abnormalities in schizophreniaand affective disorders. In C. Johnston & F. Pirozzolo (Eds.),The neuropsychology of eye movements, pp. 115—145. Hilisdale,NJ: Erlbaum.lacono, W., & Lykken, D. (1981). Two year retest stability ofeye-tracking performance and a comparison of electro-oculographicand infrared recording techniques. Psychophysiology, j., 49-55.lacono, W., Moreau, M., Beiser, M., Fleming, 3., & Lin, T.(1992). Smooth pursuit eye tracking in first epeisode psychoticpatients and their relatives. Journal of Abnormal Psychology,121, 104—16.lacono, W., Smith, N., Moreau, M., Beiser, M., Fleming, J., Lin,T., & Flak, B. (1988). Ventricular and sulcal size at onset ofpsychosis. American Journal of Psychiatry, 820—24.Illowsky, B., Juliano, D., Bigelow, L., & Weinberger, D. (1988).Stability of CT scan findings in schizophrenia: Results of an 8-year follow-up study. Journal of Neurology, Neurosurgery, andPsychiatry, j, 209—213.Jacobs, A., & Meyers, 3., (1976). Recent life events and acuteschizophrenic psychosis: A controlled study. Journal of Nervousand Mental Disease, 162, 75-87.205Jacoby, R., & Levy, R. (1980). Computed tomography in theelderly: 3. Affective disorders. British Journal of Psychiatry,iJ., 270—75.Jeste, D., Kleinman, J.., Potkin, S., Luchins, D., & Weinberger,D. (1982), Ex Uno Multi: Subtyping the schizophrenic syndrome.Biological Psychiatry, 17, 199—222.Johnston, C., & Pirozzolo, F. (Eds.) (1988). The neuropsychologyof eve movements. Hilisdale, NJ: Eribaurn.Johnstone, E., Crow, T., Frith, C., Husband, J., & Kreel, L.(1976). Cerebral ventricle size and cognitive impairment inschizophrenia. Lancet, ii, 924—926.Kaiser, M. (1974). An index of factorial simplicity.Psychometrika, 39, 31—36.Katsanis, J., lacono, W., & Beiser, M. (1991). The relationshipof lateral ventricle size to psychophysiological measures andshort-tern outcome. Psychiatry Research, 37, 115-29.Kemali, D., Mai, M., Galderisi, S., Ariano, M., Cesarelli, M.,Milici, N., Salvati, A., Valente, A., & Volpe, M. (1985).Clinical and neuropsychological correlates of cerebralventricular enlargement in schizophrenia. Journal of PsychiatricResearch, ), 587-96.Kemali, D., Maj, M., Galderisi, S., Salvati, A., Starace, F.,Valente, A., & Pirozzi, R. (1987). Clinical, biological, andneuropsychological features associated with cerebral ventricularenlargement in DSM-III schizophrenic disorder. PsychiatryResearch, II, 137—49.Kufferle, B., Grunberger, J., Linzmayer, L., & Saletu, B. (1988).Distinction between positive and negative schizophrenia in thelight of psychophysiological measurements. SchizophreniaResearch, 1, 179-180.Lazarus, R., & Folkman, S. (1984). Stress, appraisal and coping.New York: Springer.Leff, J.., Hirsch, S., Gaind, R., Rhode, P., & Stevens, B. (1973).Life events and maintenance therapy in schizophrenia. BritishJournal of Psychiatry, 123, 659—60.Leff, J., Kuipers, E., Berkowitz, R., Vaughn, C., & Sturgeon, D.(1983). Life events, relatives’ expressed emotion andmaintenance neuroleptics in schizophrenic relapse. PsychologicalMedicine, , 799—806.Leff, J., & Vaughn, C. (1985). Expressed emotion in families.New York: Guilford.206Lieberman, M. (1985). Social supports——The consequences ofpsychologizing. Journal of Consulting and Clinical Psychology,j, 461—65.Liddle, P. (1987). The symptoms of chronic schizophrenia: A reexamination of the postitive—negative dichotomy. British Journalof Psychiatry, jL, 145-51.Lorei, T., & Guel, L. (1973). Demographic characteristics aspredictors of post—hospital employment and readmission. Journalof Consulting and Clinical Psychology, jQ, 426-430.Luchins, D., Lewine, R., & Meltzer, H. (1984). Lateralventricular size, psychopathology, and medication response in thepsychoses. Biological Psychiatry, , 29-44.Luchins, D., & Meltzer, H. (1986). A comparison of CT findingsin acute and chronic ward schizophrenics. Psychiatry Research,12.., 7—14.MacMillan, J., Crow, T., Johnson, A., & Johnstone, E. (1986).The Northwick Park study of first—episode schizophrenia: III.Short—term outcome in trial entrants and trial—eligible patients.British Journal of Psychiatry, jj, 128-33.Mathew, R., Parain, C., Prakash, R., Kulkarni, M., Logan, T., &Wilson, W. (1985). A study of the septum pellucidum and corpuscallosum in schizophrenia with MR imaging. Acta PsychiatricaScandinavica, , 414-21.Matthyse, S., Holzman, P., & Lange, K. (1986). The genetictransmission of schizophrenia: Application of Mendelian latentstructure analysis to eye-tracking dysfunction in schizophreniaand affective disorder. Journal of Psychiatric Research, , 57—65.McGlashan, T. (1986). Predictors of shorter-, medium-, andlonger-term outcome in schizophrenia. American Journal ofPsychiatry, Ui, 50-55.McGlashan, T. (1988). A selective review of recent NorthAmerican long—term follow—up studies of schizophrenia.Schizophrenia Bulletin, j, 515—542.Medalie, J., & Goldbourt, U. (1976). Angina pectoris among10,000 men. American Journal of Medicine, EQ., 910—921.Meehl, P. (1962). Schizotaxia, schizoptypy, schizophrenia.American Psychologist, 17, 827—38.Mueller, D. (1980). Social networks: A promising direction forresearch on the relationship of the social environment topsychiatric disorder. Social Science and Medicine, 147—161.207Murphy, H., & Raman, A. (1971). The chronicity of schizophreniain indigenous tropical peoples: Results of a 12 year follow—upsurvey in Mauritius. British Journal of Psychiatry, JJ&, 489-497.Nasrallah, H., Kuperman, S., Hamra, B., & McKalley-Whitters, M.(1983). Clinical differences between schizophrenic patients withand without large cerebral ventricles. Journal of ClinicalPsychiatry, 44, 407—409.Nasrallah, H., McCalley—Whitters. M., & Jacoby, C. (1982).Cortical atrophy in schizophrenia and mania: A comparativestudy. Journal of Clinical Psychiatry, j, 439-441.Neale, J., & Oltmanns, T. (1980). Schizophrenia. New York:Wiley.Nuckolls, K., Cassel, J., & Kaplan, B. (1972). Psychosocialassets, life crises and the prognosis of pregnancy. AmericanJournal of Epidemiology, 95, 431—441.Nuechterlein, K. (1987). Vulnerability models for schizophrenia:State of the art. In H. Hafner, W. Gattaz, & W. Janzarik (Eds.),Search for the causes of schizophrenia, pp. 297-316. New York:Springer.Nuechterlein, K., & Dawson, M. (1984). A heuristicvulnerability! stress model of schizophrenic episodes.Schizophrenia Bulletin, Q., 300—312.Olsen, S., Nasrallah, H., Coffman J., & Swartzkopf, S. (1990).CT and MRI abnormalities in schizophrenia: Relationship withnegative symptoms. In J. Greden & R. Tandon (Eds.), Negativeschizophrenic symptoms: Pathophysiology and clinicalimplications, pp. 147-60. Washington, DC: American PsychiatricAssociation Press.Owen, M., Lewis, S., & Murray, R. (1989). Family history andcerebral ventricular enlargement in schizophrenia: A casecontrol study. British Journal of Psychiatry, jj, 629—34.Owens, D., Johnstone, E., Crow, T., Frith, C., Jagoe, J., &Kreel, L. (1985). Lateral ventricle size in schizophrenia:Relationship to the disease process and its clinicalmanifestations. Psychological Medicine, , 27—41.Pandurangi, A., Bilder, R., Rieder, R., Mukherjhee, S., & Hamer,R. (1988). Schizophrenic symptoms and deterioration in relationto computed tomographic findings. Journal of Nervous and MentalDisease, 200—206.Pattison, E., DeFrancisco, D., Wood, P., Frazier, H., & Crowder,J. (1975). A psychosocial kinship model for family therapy.American Journal of Psychiatry, 1246—1251.208Pearison, G., Garbacz, D., Breakey, W., Ahn, H., & DePaulo, .J.(1984). Lateral ventricular enlargment associated withpersistent unemployment and negative symptoms in bothschizophrenia and bipolar disorder. Psychiatry Research, , 1-9.Pearlson, G., Garbacz, D., Moberg, P., Ahn, H., & DePaulo, J.(1985). Symptomatic, familial, perinatal, and social correlatesof computerized axial tomography (CAT) changes in schizophrenicsand bipolars. Journal of Nervous and Mental Disease, 131, 42-50.Penn, R., Belanger, M., & Yasnoff, W. (1978). Ventricular volumein man computed from CAT scans. Annals of Neurology, , 216-223.Pfefferbaum, A., Zipursky, R., 1dm, K., Zatz, L., Stahl, S., &Jernigan, T. (1988). Computed tomographic evidence forgeneralized sulcal and ventricular enlargement in schizophrenia.Archives of General Psychiatry, j., 633—640.Porrit, D. (1979). Social support in crisis: Quantity orquality? Social Science and Medicine,.(6A), 715—721.Prudo, R., & Blum, H. (1987). Five year outcome and prognosis inschizophrenia. British Journal of Psychiatry, jQ., 345—354.Raz, S., & Raz, N. (1990). Structural brain abnormalities in themajor psychoses: A quantitative review of the evidence fromcomputerized imaging. Psychological Bulletin, j.Q, 93-108.Reddy, R., Mukherjee, S., Schnur, D., Chin, J., & Degreef, G.(1990). History of obstetrical complications, family history,and CT scan findings in schizophrenic patients. SchizophreniaResearch, j, 311-14.Reveley, A., Reveley, M., Clifford, C., & Murray, R. (1982).Cerebral ventricular size in twins discordant for schizophrenia.Lancet, , 540—541.Rieder, R., Mann, L., Weinberger, D., van Kammen, D.,, & Post, R.(1983). Computed tomographic scans in patients withschizophrenia, schizoaffective, and bipolar affective disorder.Archives of General Psychiatry, jQ, 735—739.Robins, L., Helzer, J., Croughan, J., & Ratcliff, K. (1981). TheNIMH Diagnostic Interview Schedule: Its history, characteristicsand validity. Archives of General Psychiatry, , 381-389.Roy—Byrne, P., Post, R., Keilner, C., Joff, T., & Vhde, T.(1988). Ventricular—brain ratio and life course of patients withaffective disorder. Psychiatry Research, , 277-284.209Sacchetti, E., Caizeroni, A., Vita, A., Terzi, A., Pollastro, F.,& Cazzulo, C. (1992). The brain damage hypothesis of theseasonality of births in schizophrenia and major affectivedisorders: Evidence from computerized tomography. BritishJournal of Psychiatry, 160, 390—97.Schradle, S., & Dougher, M. (1985). Social support as a mediatorof stress: Theoretical and empirical issues. ClinicalPsychology Review, , 641—661.Schulz, S., Koller, M., Kishore, P., Hamer, R., Gehi, 3., &Friedel, R. (1983). Ventricular enlargement in teenage patientswith schizophrenia spectrum disorder. American Journal ofPsychiatry, 140, 1592—1595.Scott, M., Golden, C., Ruedrich, S., & Bishop, R. (1983),Ventricular enlargement in major depression. PsychiatryResearch, , 91—93.Seeman, M. (1982). Gender differences in schizophrenia.Canadian Journal of Psychiatry, , 107-112.Seidman, L. (1983). Schizophrenia and brain dysfunction: Anintegration of recent neurodiagnostic findings. PsychologicalBulletin, j, 195—238.Seidman, L., Sokolove, R., McElroy, C., Knapp, P.., & Sabin, T.(1987). Lateral ventricular size and social networkdifferentiation in young, nonchronic schizophrenic patients.American Journal of Psychiatry, jj, 512—514.Shagass, C., Roemer, R., & Amadeo, M. (1976). Eye-trackingperformance and engagement of attention. Archives of GeneralPsychiatry, 33, 121-125.Shaw, R. (1985). Intermetropolitan migration in Canada.Toronto: NC Press.Shima, S., Shikano, T., Kitamura, T., Masuda, Y., Tsukumo, T.,Kanba, S., & Asai, M. (1984). Depression and ventricularenlargement. Acta Psychiatrica Scandinavica, IQ., 275-77.Siever, L., Coursey, R., Alterman, I., Buchsbaum, M., & Murphy,D. (1984). Impaired smooth pursuit eye movement: Vulnerabilitymarker for schizotypal personality disorder in a normal volunteerpopulation. American Journal of Psychiatry, jJi, 1560-1566.Siever, L., van Kammen, D., Linnoila, M., Alterman, I., Hare, T.,& Murphy, D. (1986). Smooth pursuit eye movement disorder andits psychobiologic correlates in unmedicated schizophrenics.Biological Psychiatry, 2J, 1167—1174.210Smith, G. (1986). A computerized tomographic evaluation of brainmorphology in first episode psychotic patients. Unpublisheddoctoral dissertation, University of British Columbia, Vancouver.Smith, R., & Maser, J. (1983). Morphological andneuropsychological abnormalities as predictors of clinicalresponse to psychotropic drugs. Psychopharmacological Bulletin,.i, 505—09.Sokolovsky, 3., Cohen, C.., Berger, D., & Geiger, J. (1978).Personal networks of ex-mental patients in a Manhattan SRO hotel.Human Organization, 3, 5-15.Sponheim, S., lacono, W., & Beiser, M. (1991). Stability ofventricular size after the onset of psychosis. PsychiatryResearch, 40, 21-29.Spring, B. (1981). Stress and schizophrenia: Some definitionalissues. Schizophrenia Bulletin,,24—33.Stephens, J. (1978). Long—term prognosis and follow-up inschizophrenia. Schizophrenia Bulletin, ,j, 25—47.Stoffelmayr, B., Dillavou, D., & Hunter, J. (1983). Premorbidfunctioning and schizophrenia. Journal of Consulting andClinical Psychology, j, 338—352.Strauss, J., & Carpenter, W. (1977). The prediction of outcomein schizophrenia: III. Five year outcome and its predictors.Archives of General Psychiatry, 34, 159—163.Sturgeon, D., Turpin, G., Kuipers, L., Berkowitz, R., & Leff, .3.(1984). Psycophysiological responses of schizophrenic patientsto high and low expressed emotion relatives: A follow-up study.British Journal of Psychiatry, 145, 62—69.Suddath, R., Christison, G., Torrey, E., & Weinberger, D. (1990).Anatomical abnormalities in the brains of monozygotic twinsdiscordant for schizophrenia. New England Journal of Medicine,aaz., 789—794.Szymanski, S., Kane, 3., & Lieberman, J. (1991). A selectivereview of biological markers in schizophrenia. SchizophreniaBulletin, U, 99-111.Takahashi, R., manage, V., & Kato, N. (1982). CT scanning andthe investigation of schizophrenia. In C. Perris, G. Struwe, &B. Jansson (Eds.), Biological Psychiatry, pp. 259-68.Targum, S., Rosen, L., DeLisi, L., Weinberger, D., & Citrim, C.(1983). Cerebral ventricular size in Major Depressive disorder:Association with delusional symptoms. Biological Psychiatry,,329—336.211TerBrugge, K., & Rao, K. (1983). Hydrocephalus and atrophy. InS. Lee & K. Rao (Eds.), Cranial computed tomography, 170-199.Toronto: McGraw Hill.Thoits, p. (1983). Dimensions of life events that influencepsychological distress: An evaluation and synthesis of theliterature. In H. Kaplan (Ed.), Pychosocial stress: Trends intheory and research, pp. 33-103. New York: Academic.Tolsdorf, C. (1976). Social networks, support, and coping: Anexploratory study. Family Process, j, 407—417.Turner, R. (1983). Direct, indirect, and moderating effects ofsocial support on psychological distress and associatedconditions. In H. Kaplan (Ed.), Pychosocial stress: Trends intheory and research, pp. 105-55. New York: Academic.Turner, R., & McLean, p. (1989). Physical disability andpsychological distress. Rehabilitation Psychology, jj, 225—242.Turner, S., Toone, B., & Brett—Jones, J. (1986). Computedtomographic changes in early schizophrenia: Preliminaryfindings. Psychological Medicine, , 219—26.Vaillant, G. (1964). Prospective prediction of schizophrenicremission. Archives of General Psychiatry, 11, 509—518.Ventura, J., Nuechterlein, K., Lukoff, D., & Pederson-Hardesty,J. (1989). A prospective study of stressful life events andschizophrenic relapse. Journal of Abnormal Psychology, , 407-411.Vita, A., Saccetti, E., Valvassori, G., & Cazzullo, C. (1988).Brain morphology in schizophrenia: a 2- to 5-year CT scan follow-up study. Acta Psvchiatrica Scandinavica, Z., 618-621.Waxler, N. (1977). Is outcome for schizophrenia better innonindustrial countries? Journal of Nervous and Mental Disorder,167, 144—158.Weinberger, D. (1987). Implications of normal brain developmentfor the pathogenesis of schizophrenia. Archives of GeneralPsychiatry, jj, 660-669.Weinberger, D., Cannon—Spoor, E., Potkin, S., & Wyatt, R. (1980).Poor premorbid adjustment and CT scan abnormalities in chronicschizophrenia. American Journal of Psychiatry, i, 1410—1413.Weinberger, D., DeLisi, L., Perman, G., Targum, S., & Wyatt, R.(1982). Computed tomography in schizophreniform disorder andother psychiatric disorders. Archives of General Psychiatry, ,778—783.212Weinberger, D., DeLisi, L., Neophytides, A., & Wyatt, R. (1979).Familial aspects of CT scan abnormalities in chronicschizophrenic patients. Psychiatry Research,,65-71.Weiss, R. (1974). The provisions of social relationships. In Z.Rubin (Ed.), Doing unto others, pp. 17—26. Englewood Cliffs,NJ: Prentice-Hall.Westermeyer, J., & Pattison, E. (1981). Social networks andmental illness in a peasant society. Schizophrenia Bulletin, 7,125—134.Williams, A., Reveley, M., Kolakowksa, T., Ardern, M., &Mandelbrote, B. (1985). Schizophrenia with good and poor outcomeII.: Cerebral ventricle size and its clinical significance.British Journal of Psychiatry, Jj.., 239-46.Wing, J., Cooper, J., & Sartorius, N. (1974). The measurementand classification of psychiatric symptoms. London: CambridgeUnversity Press.World Health Organization (1972). The international pilot studyon schizophrenia. Geneva: Author.World Health Organization (1978). Mental disorders: Glossaryand guide to their classification in accordance with the ninthrevision of the International Classification of Diseases.Geneva: Author.World Health Organization (1979). Schizophrenia: Aninternational follow-up study. New York: Wiley.Zubin, J., & Spring, B. (1977). Vulnerability: A new view ofschizophrenia. Journal of Abnormal Psychology, j, 103-126.Zubin, J., & Steinhauer, S. (1983). The metamorphosis ofschizophenia: From chronicity to vulnerablity. PychologicalMedicine, , 551—71.213Arjvendix 1.Summary of DSM-III Axis VLevels Adult examples Child/Adol. examplesSUPERIOR--Unusually effective functioning in socialrelations, occupational functioning,& use of leisuret i me.Single parent livingin deterioratingneighbourhood takesexcellent care ofchildren & home, haswarm relations withfriend, & finds timeto pursue hobby.A 12—year—old girlgets superior gradesin school, is extremely popular amongpeers, & excels inmany sports. Shedoes all of thiswith ease & comfort.VERY GOOD--Betterthan average functioning in socialrelations, workfunctioning, & useof leisure time.A 65—yr.—old retiredwidower does somevolunteer work,often sees oldfriends, & pursueshobbies.Adolescent boy getsexcellent grades,works part—time, hasseveral closefriends & playsin jazz band.GOOD--No more thanslight impairmentin either social oroccupationalfunctioning.A woman with manyfriends functionswell at a difficult job, but saysstrain is too much.An 8-year-old boydoes well at school,has several friends,but bulliesyounger children.FAIR--Moderate impairment in eithersocial or occupational functioning,or some impairmentin both.A lawyer has trouble carrying throughassignments; has several acquaintancesbut hardly any closefriends.A 10—year—old girldoes poorly inschool but has adequate peer and familyrelations.POOR--Marked impairment in eithersocial relations oroccupational functioning, or moderateimpairment in both.A man with one ortwo friends hastrouble keeping ajob for more thanfew weeks.A 14-year-old boyalmost fails inschool & has troublea getting along withhis peers.VERY POOR--Markedimpairment in bothsocial relations &occupational functtioning.A woman is unable todo any of her housework & has violentoutbursts toward herneighbours.A 6—year—old girlneeds special helpin all subjects &has virtually nopeer relationships.GROSSLY IMPAIRED--Gross impairment invirtually all areasof functioning.An elderly man needssupervision to maintain minimal hygieneand is usuallyincoherent.A 4-yr-old boy needsconstant restraintto avoid self-injury& is almost totallylacking in socialskills.


Citation Scheme:


Citations by CSL (citeproc-js)

Usage Statistics



Customize your widget with the following options, then copy and paste the code below into the HTML of your page to embed this item in your website.
                            <div id="ubcOpenCollectionsWidgetDisplay">
                            <script id="ubcOpenCollectionsWidget"
                            async >
IIIF logo Our image viewer uses the IIIF 2.0 standard. To load this item in other compatible viewers, use this url:


Related Items