The following projects are either ongoing or proposed. Not all projects on the list will necessarily be completed. This page is for general interest only.

 

• PROJECT 1: A PROSPECTIVE, LONGITUDINAL STUDY OF RECENT ONSET PSYCHOSIS
• PROJECT 2: A PROSPECTIVE LONGITUDINAL EVALUATION OF COGNITIVE DYSMETRIA IN FIRST EPISODE PATIENTS SUBSTUDY 2
• PROJECT 3: EVALUATING ALTERNATIVE APPROACHES TO MAKING THE DEFICIT VS. NON-DEFICIT DISTINCTION
• PROJECT 4: EVALUATION OF PREMORBID COGNITIVE FUNCTION USING CHILDHOOD SCHOLASTIC APTITUDE TEST SCORES
• PROJECT 5: CHARACTERIZATION OF CHRONIC SCHIZOPHRENIA IN LATE LIFE: NEUROPSYCHOLOGICAL AND SYMPTOM ASSESSMENT OF ELDERLY SUBJECTS
• PROJECT 6: DEVELOPMENT OF A NEW SCALE TO MEASURE PSYCHOPATHOLOGY IN SCHIZOPHRENIA: THE "5 STEPS"

 

---

PROJECT 1: A PROSPECTIVE, LONGITUDINAL STUDY OF RECENT ONSET PSYCHOSIS
Principal Investigator: Nancy C. Andreasen, M.D., Ph.D.
Co-Investigators: Peg Nopoulos, M.D., Michael Flaum, M.D., Del Miller, Pharm.D., M.D., Jane Paulsen, Ph.D., Dan O'Leary, Ph.D., Stephan Arndt, Ph.D.
Funding Status: This project is funded principally by an RO1 grant to Dr. Andreasen, entitled "Phenomenology and Classification of Schizophrenia."
Progress Status and
Proposed Duration: This project has been ongoing for nine years. It was initially funded by the ROI grant noted above in 1987, and renewed in 1992. The grant is currently being considered for continued funding through 2002.
Specific Aim
1. To continue to explore the three dimensions of psychopathology in schizophrenia (psychoticism, disorganization, and negative symptoms) in relation to outcome.
2. To continue our prospective longitudinal follow-up of 260 patients in order to describe the long-term outcome of schizophrenia by examining four domains: symptoms, cognitive function, psychosocial function ("quality of life"), and brain morphology.
3. To further explore the predictors of outcome in this sample, including premorbid function, symptom type and severity, neurological soft signs, and neuropsychological/cognitive performance.
4. To continue our investigation of the role of developmental processes as causative or contributing factors to the occurrence of the illness.

Background and Rationale
As in any medical illness, the most important questions that the patient and family want to have answered are those involving prognosis. What is the expected course of a young person following a first episode of psychosis? Follow-up studies of long-term outcome in schizophrenia have shed some light on the issue of prognosis. They have demonstrated that patients with good premorbid functioning and acute onset tend to do better than those with long-standing problems and an insidious onset. Those who present with prominent features of anxiety, subjective distress, or confusion, appear to have a better prognosis than those in whom disorganized thought, assaultiveness, or emotional blunting dominate the clinical picture. Still, many of the basic questions concerning course and outcome of schizophrenia remain unanswered, and there is considerable disagreement between many of the larger outcome studies. The present investigation is designed as an intensive study of an informative sample: patients experiencing their first episode of schizophrenia or a schizophrenia spectrum disorder, or those early in the course of these disorders. The goal is to follow these patients prospectively for at least an 8 - 10 year period with relatively intense surveillance in order to document the clinical course in the context of a wide variety of factors. It examines the predictive validity of patterns and types of symptoms assessed cross-sectionally at intake, and supplements this comprehensive clinical information with neuropsychological and neurobiological measures, such as magnetic resonance imaging. Environmental risk factors and genetic vulnerability are also evaluated as predictors of outcome.
Hypotheses
Some of the hypotheses being explored are summarized below.

The Three Dimensions of Psychopathology
1. In contrast to psychotic and disorganized symptoms, negative symptoms will be predictive of poor outcome as assessed by psychosocial factors (e.g., employment, marital status) and measures of psychiatric morbidity (e.g., duration of hospitalization, severity of symptoms).
2. The three domains of psychopathology will remain independent of each other over time with positive and disorganized symptoms waxing and waning while negative symptoms remain mostly stable.
3. Negative symptoms will be associated with cognitive dysfunction, brain abnormalities, and neurologic soft signs.

Course and Outcome
1. Structural brain abnormalities will be present early in the illness and will remain stable in most cases.
2. Premorbid cognitive function as determined by childhood scholastic aptitude test scores (e.g., the Iowa Test of Basic Skills) will be abnormal in preschizophrenic children as early as the 4th grade, and will worsen during the 1-2 years prior to the onset of psychotic symptoms.
3. Cognitive dysfunction at the time of diagnosis will be significant, but will remain stable thereafter, with even some mild long-term improvement.

Predictors of Outcome
1. Measures suggestive of abnormal development such as birth complications, poor premorbid cognitive function, and poor childhood social adjustment will be associated with poor outcome.
2. Patients who receive early treatment and who comply with medications will have significantly better psychosocial and psychiatric outcome.
3. Prominent negative symptoms present early in the course of illness will be associated with poorer outcome.
4. Insight into illness will be related to suicidality and attempted/completed suicides.

Neurodevelopmental vs. Neurodegenerative Pathophysiology
1. Indicators of abnormal development such as birth complications, poor premorbid cognitive function and poor childhood social adjustment will be associated with measures of neural abnormalities such as abnormal brain morphology, soft signs and abnormal movements, and cognitive dysfunction.
2. Age of onset of illness will be correlated with severity and/or number of measures of abnormal development. That is, those patients with several indicators of abnormal development will have earlier age of onset.
3. The majority of patients will show some evidence of abnormal development (e.g., premorbid cognitive dysfunction, premorbid social dysfunction, brain morphologic changes present at onset of illness).
4. A minority of patients will show significant deterioration in function over time as evidenced by increased CSF, decreased brain tissue, and worsening cognitive function, suggesting a process of neurodegeneration.
5. Those patients affected most by abnormal development will be the most vulnerable to deterioration. Therefore, the patients who show the most evidence of neurodegeneration will be those patients who rate the highest in abnormal development.

Methods
Patient Sample: The patient sample consists of two related groups. One group is a first episode sample, consisting of subjects who are experiencing a psychosis severe enough to require hospitalization for the first time (current n =124). The second group consists of "recent onset" patients (current n = 134). These are patients who have had five years or less of illness as dated from the time of the first hospitalization. Since these patients are relatively young and have been ill only for a short time, most have not accumulated a substantial amount of psychosocial morbidity as a consequence of the illness. Thus the evolution of their symptoms and prognosis can also be studied effectively in the prospective design proposed.

Included in this sample are not only patients with the diagnosis of schizophrenia, but also those diagnosed as having a mood disorder with mood incongruent psychotic features. We have decided to include these patients for two reasons. First, the boundary between this diagnostic category and the schizophrenia spectrum remains unclear and requires validation. We suspect that the current definition of the schizophrenia spectrum (as per DSM criteria) may prove to be unrealistically narrow. Second, we have observed that in questionable cases, patients are often given the "benefit of the doubt" and assigned the diagnosis with the better prognosis and more benign treatment (i.e., mood disorder rather than schizophrenia). Although such practice is understandable in a clinical setting, for the purposes of a prospective longitudinal study, it may result in a selection bias in which schizophrenic patients who initially present with mood symptoms as well as psychotic symptoms are excluded.
Predictors of Outcome: The basic design of the study involves attempting to identify useful predictors of outcome. Based on clinical experience and previous research, we have identified a set of predictors that we expect to be potentially useful.

Predictor Variables
Outcome Variables
Gender Time to recovery (or presence of chronicity)
Age of onset Time to relapse
Sociodemographic measures Number of relapses
(e.g., age, education, current employment) Number of hospitalizations

Types of symptoms
Type of symptoms
(e.g., positive, negative, affective)
Severity of symptoms
Premorbid personality adjustment Social adjustment
Brain morphology (e.g., employment, social contacts)
Neurologic soft signs Cognitive function
Birth and developmental history Movement abnormalities
(e.g., prematurity, prolonged labor, Suicidality
delayed milestones, etc.)
Family History
Cognitive function
Insight
We recognize that a design involving multiple predictors and multiple outcomes is a complicated one. Yet, this design reflects the genuine complexity of prediction for the schizophrenic syndrome. We have attempted to simplify the design in part by making specific directional hypotheses concerning individual predictors in isolation.
Measurements and Assessments at Intake: Subjects in this study are assessed with the full complement of phenomenological and neurobiological measures, most of which are described in detail in the Assessment and Training Core of this application. A comprehensive assessment of type and severity of psychiatric symptoms and psychosocial functioning both presently and in the past is gathered through the CASH and the PSYCH. Family history is documented via the FH-RDC (Family History - Research Diagnostic Criteria) and birth history via the format developed by DeLisi. The cognitive battery is also described in that unit. We are also now beginning to collect Childhood Home Movies to evaluate affect and motor abnormalities. The protocol for magnetic resonance imaging is detailed in the Structural Imaging Unit. The neurological exam includes measures of localizing signs, "soft signs," and abnormal developmental reflexes. The AIMS (Abnormal Involuntary Movements Scale) is used for the assessment of tardive dyskinesia. The medical history checklist documents the history and treatment of any serious medical/surgical disorders. Diagnoses are made using all sources of information available and a consensus diagnostic conference where at least three research psychiatrists are present. This conference is also videotaped, providing for yet another assessment instrument to evaluate longitudinally.

Follow-up Assessments: All patients entered into the study are subsequently evaluated at regular 6-month intervals thereafter. The evaluations obtained at each of the 6-month intervals focus on type and severity of symptoms, treatment, and psychosocial function; the CASH-UP and PSYCH-UP are the primary evaluation instruments. At two, five, and nine years, we obtain more comprehensive evaluations. Additional measures at these points include the neurologic exam, AIMS, cognitive battery, MR, consensus diagnostic conference (see also Assessment and Training Core for details), and video.
Progress Report: Many of the findings from this sample are outlined above in this unit's overall progress report. Below is a summary of the progress of the sample collection.

The original grant proposed to collect 150 subjects (60 first episode and 90 recent onset), and the number was projected to double when the grant period was extended for another five years through the MERIT award mechanism, to achieve a total of approximately 300. Intake has in fact exceeded expectations in all respects, with a total enrollment of 325 subjects, nearly 260 of which we are now actively following. Further, the study has gained momentum as our CRC has matured and as physicians throughout the state have become increasingly aware of our interest in studying first episode and neuroleptic naive patients and have developed patterns of referring them to us. Enrollment has grown substantially since 1994, and our proportion of first episode and neuroleptic naive patients has increased. Further, our attrition rate in the follow-up component (i.e., refused, lost, suicide, death) has decreased. Because of its stable population base, as well as a tradition of civic responsibility that includes both family members and patients, Iowa is an ideal place for a prospective longitudinal study of the type we are conducting. This is also a relatively "clean" sample of largely community-based patients with relatively intact families and very low rates of co-morbid substance abuse, permitting us to study the longitudinal course of symptoms and psychosocial outcome in schizophrenia uncontaminated by confounds such as institutionalization, poverty and malnutrition, or substance abuse.

The cohort of nearly 260 patients that we are currently actively following constitute an extremely valuable resource for determining the symptomatic and anatomic evolution of schizophrenia over time. This sample is just now "maturing," in that a large number of subjects are in the cohort that has been followed for more than two years and will be returning for five and nine year follow-ups during the upcoming funding period. For example, 150 subjects will have a five-year follow-up, while 110 will have a nine-year follow-up within the next five years. Since MR and cognitive assessments are obtained at these time points, we should be able to achieve a relatively definitive resolution of the developmental vs degenerative question, due to the availability of this large sample of "clean" patients, many of whom entered the study in their first episode and while neuroleptic naive. In addition, this is clearly a valuable sample to determine the naturalistic outcome of schizophrenia and to identify the predictors of outcome (e.g., depressive symptoms, insight). For example, eleven of these prospectively ascertained patients have completed suicide, and the N in this informative subgroup will (sadly) continue to increase, permitting us to identify the factors that put these young acute patients at risk for suicide and perhaps suggest improved interventions.
Data Analysis
Most of the analyses pertaining to the prediction of outcome are relatively straightforward forms of bivariate and multiple regression where outcome is defined as a continuous measure (e.g., severity of symptoms at a target period). Other regression-like analyses will be used as appropriate such as logistic regression models when outcome is expressed in some other fashion (e.g., binary data such as 'Was the patient rehospitalized within the year?'), or parametric and nonparametric survival analysis when time to episode or rehospitalization is of interest. The presence of two independent samples allows us to perform replications in addition to standard statistical tests. Also, tests of similarity across diagnoses (e.g., schizophrenic, schizoaffective, schizotypal, and mood disorder) will be included in our analyses.

While standard types of regression analyses will be of value for a number of hypotheses, the richness of our longitudinal data will allow for other interesting kinds of questions pertaining to a comprehensive description of the evolution of phenomenology. Cross - lagged panel (Kenny et al 1979) and other path type analyses can be used to address the relative contribution of symptoms at an index period to the development of subsequent symptom patterns. Furthermore, in describing the course of symptom data a host of recently developed techniques may be applied for estimating and depicting developmental models. Dr. R. F. Woolson (Department of Preventive Medicine and Environmental Health) and Dr. S. Arndt, among others, have been awarded a grant studying the properties of recently developed methods for analyzing longitudinal data when some of the data are incomplete (i.e., due to missing observations), further developing semi-parametric methodology (e.g., methods developed by Stram and Landis) for analyzing repeated measures of ordinal scale response variables, developing goodness of fit procedures for assessing the adequacy of longitudinal models, and nonparametric methods for comparing groups of subjects when the response variable is non-normal.

Analyses verify that these samples (first episode and recent onset n's of approximately 130 each) have excellent power to detect the associations of interest. For instance a realistic analysis would use multiple regression with 4 covariates and one independent variable of interest. If the covariates accounted for 50 percent of the variance, the test for a single independent variable accounting for only 5 percent of the variance will have a power of 0.967 using a 0.05 significance threshold (alpha), and a power of 0.87 using the more stringent 0.01 alpha. Assuming equal R2 values for the variable of interest in both populations, the chance of duplicating the result using the two-sample replication strategy is power > 0.92 using the 0.05 alpha.

Back to top

PROJECT 2: A PROSPECTIVE LONGITUDINAL EVALUATION OF COGNITIVE DYSMETRIA IN FIRST EPISODE PATIENTS SUBSTUDY 2
Principal Investigator: Nancy C. Andreasen, M.D., Ph.D.
Co-Investigators: Peg Nopoulos, M.D., Michael Flaum, M.D., Del Miller, Pharm.D., M.D., Dan O'Leary, Ph.D., Jane Paulsen, Ph.D., Stephan Arndt, Ph.D.
Funding Status: This project is funded principally by an RO1 grant to Dr. Andreasen, entitled, "Phenomenology and Classification of Schizophrenia"
Progress Status and Proposed Duration: This project has been ongoing for nine years. It was initially funded by the ROI grant noted above in 1987, and renewed in 1992. The grant is currently being considered for continued funding through 2002.

Specific Aims
1. To develop new approaches to assessing the phenomenology of schizophrenia, with an emphasis on identifying and measuring abnormalities in basic cognitive processes that may lead to its heterogeneity of symptoms, particularly cognitive dysmetria.
2. To collect a new cohort of first episode patients who will be assessed with these measures in a prospective longitudinal design while in a standard treatment protocol utilizing atypical neuroleptics.
3. To apply improved or updated methods for these assessments whenever possible (e.g., cognitive measures targeted to cognitive dysmetria).
4. To examine long-term outcome in this new sample, assessing the same four domains that have been explored in our current cohort (symptoms, cognitive function, psychosocial function, and brain morphology).
5. To explore predictors of outcome in this sample, employing improved measures of premorbid function (e.g., school records, home movies), neurological/motor function, and cognitive/neuropsychological performance.
Background and Rationale
This study is an extension of Project 1 which is a continued follow-up of first episode and recent onset patients that we are currently following. Longitudinal studies, by their very nature, can become rigid and outdated, since they become wedded to maintaining a specific set of measures in a world where the state of the art is changing. We, in fact, believe that the comprehensive measures selected for Project 1, described above, were sufficiently forward-looking that they are still relatively "young" and are in fact the types of assessments that are being used in studies that are being initiated at the present time.

Given the opportunity to "update" the study in the recent submission for continued funding of this grant, we concluded that the "ideal design" would permit us to continue to collect the assessments used to date, but also to update the design by adding several new measures that will permit us to examine the longitudinal course of schizophrenia and predictors of outcome in a twenty-first century style.

Consequently, the patients in this new cohort will be assessed with all of the scales, procedures, and instruments as outlined for Project 1 above. This will provide continuity with the work completed to date and continue to enlarge our sample, particularly the highly informative first episode group.

In addition, however, the patients in Project 2 will also be given new and more sophisticated assessments, targeted to measure the primary cognitive deficit, dysmetria. These additional assessments draw on techniques of experimental cognitive psychology, as an addition to the more traditional neuropsychological techniques of Project 1. Further, the patients in this sample will also be assessed with a newly developed MR sequence that permits a more fine-grained assessment of brain morphology, but which is compatible with the MR methods used in Project 1.

Hypotheses
1. Severity of cognitive dysmetria will be related to poor outcome as measured by severity of symptoms and impairment in psychosocial functioning.
2. Measures of cognitive dysmetria, as assessed in neuroleptic naive patients and again after treatment with atypical neuroleptics, will show improvement. Degree of improvement in dysmetria will be related to degree of clinical improvement.
3. Severity of cognitive dysmetria will be positively correlated to severity of abnormalities in brain morphology.
4. Cognitive dysmetria will be present and severe at first presentation of illness (first episode patients) and will remain stable over time in most patients with some modest improvement related to clinical improvement.
5. The "new cohort," treated exclusively with atypical neuroleptics, will have a better outcome than the "old cohort" that was treated with typicals, when the same time periods are compared with one another (e.g., first five years after initial presentation).

Methods
Patient Sample: This group of patients will be limited to first episode patients. We have established the ability to recruit these patients at a rate of about 20-25 per year. Many of these patients will be neuroleptic naive, which will minimize the confounding effects of chronicity or effects of treatment which are areas of considerable concern for a hypothesis about motor and cognitive coordination.

New and Improved Assessments: The tasks to be used assess specific aspects of the processes that are involved in the fluid coordination of mental functions, starting with attention and early sensory processing and assessing cognitive processes that affect motor function in a fine-grained fashion. These tasks are described in detail in the Cognitive Neuroscience Section.

Classical Eyeblink Conditioning: Classical eyeblink conditioning involves presentation of a conditioned stimulus (CS, e.g., a tone) in close temporal proximity to an unconditioned stimulus (US e.g., air puff to eye) that elicits an unconditioned response (US, an eyeblink). Through repeated pairings in which the CS precedes the US by a small interval (optimally 200-500 msec), associative learning occurs and the CS elicits a conditioned eyeblink response (CR). Recording, lesion, stimulation and anatomical evidence from human and animal studies indicate that the cerebellum is the critical neural substrate of eyeblink conditioning.

Backward Masking: We will directly replicate the procedure used by Braff, which involves establishing critical stimulus thresholds and the critical interstimulus interval (ISI) for backward masking using an "A" versus a "T" as a stimulus. We will also investigate critical stimulus durations and backward masking using more complex stimuli.

Facilitation and Inhibition of Motor Responses: The Flanker task is a choice reaction time paradigm in which a subject is required to respond when one of several possible target letters (e.g., an X or an O) appears in the center of a letter display. Reaction time is slowed when the central target letter is surrounded or flanked by letters that are potential targets but are not in the center of the display (e.g., an X flanked by O's). In contrast, reaction time is faster when the central target letter is flanked by letters that require the same response as the center letter (e.g., an X flanked by X's).

Time Perception and Time Production: We will explore timing functions using tasks described in Ivry and Keele, who studied patients with lesions of different types (Parkinson, cerebellar, cortical and peripheral neuropathy) on two measures of timing function. One task involves the production of timed intervals in which subjects try to maintain a simple rhythm. The other task measures a subject's ability to perceive small differences in the duration of two intervals.

Procedural Learning and Memory: Initiation and Maintenance of Voluntary Motor Acts: We will explore the initiation of motor action using a task described by Malenka et al. The task requires the subject to track a target (a moving spot of light on an oscilloscope screen) by moving a joystick. The joystick also moves a spot of light on the screen, but the relationship between the direction of movement of the joystick and the direction of movement of the tracking spot can be manipulated by the experimenter.

Prism Adaptation: Throwing an object at a target requires integration of visual and motor information. The relationship between vision and arm movements is adjustable, as can be demonstrated by the adaptation of normal subjects to wedge prisms which displace visual input. Throwing while looking through prisms: I. Focal olivocerebellar lesions impair adaptation. Brain, 119:1183-1198; Martin, T.A., J.G. Keating, H.P. Goodkin, A.J. Bastian and W.T. Thach (1996). Throwing while looking through prisms: II. Specificity and storage of multiple gaze-throw calibrations will be used as a functional probe of the intactness of discrete regions of the cerebellar system in patients with schizophrenia.

Brain Morphology and the cortico-cerebellar-thalamic-cortical circuit (CCTCC): Our work to date has employed multispectral techniques (PD, T1, T2 sequences) and 1x1x1.5 mm voxels in order to extract maximal information and develop very accurate tissue classification techniques that form the basis for our surface-finding techniques and the measurement of sulcal/gyral anatomy. For the studies of the new longitudinal sample we will use a new sequence that we have developed and that is targeted to address the specific questions/regions that will give us better measurements of the three regions that are of specific interest for the study of cognitive dysmetria: the prefrontal cortex, the cerebellum, and the thalamus.

The measurements of all three of these nodes in the CCTCC will be done initially via a combination of manual segmentation and our current automated methods (e.g., atlas-based division into lobes and regions). Using image analysis techniques developed through our other RO1 grant, Brain Imaging in the Major Psychoses (e.g., neural nets, high dimensional transformations), we will also create methods that will be fully automated and therefore highly reliable and efficient (details of this project are discussed in the Structual Imaging Unit).

Standardized Treatment Protocol: We will control treatment for this new cohort of patients using a protocol involving atypical neuroleptics. The reason for controlled treatment is several-fold. First, we propose the primary deficit in patients with schizophrenia is dysfunction of the CCTC circuit, resulting in dysmetria of both cognitive and motor function. Many neuroleptics, especially the typical ones, have direct and significant effects on the motor system which may therefore cloud our ability to assess subtle, primary motor dysfunction. Secondly, the atypical neuroleptics have shown to have at least modest amounts of improved efficacy over typical neuroleptics in treatment of negative symptoms, as well as cognition and psychosocial functioning. Therefore, it is possible that the optimal treatment for these young patients is treatment with this class of medicines. Finally, since uncontrolled treatment with many types of drugs, side effects, and doses may confound the ability to define predictive variables, using a controlled treatment strategy offers the best opportunity at isolating those factors that influence course and outcome of the disease.
Data Analysis
Since many of the indices for dysmetria either involve new measures (e.g., time estimation, prism adaptation) we plan to examine their distributional and psychometric properties before choosing the exact form of regression analysis. Exploratory analyses will investigate the interrelationships and interactions of the dysmetria measures.

When continuous response data are clearly non-normal or there are other problems with the parametric assumptions, nonparametric alternatives will be used, such as regression with rank transformed data. Alternatively, simple data transformation may be applicable to normalize data. Other regression analyses will be used as appropriate, such as logistic regression models when outcome is expressed in categorical terms (e.g., need for rehospitalization, suicidality). Logistic regression will also be used when predicting remission or recovery. For instance, parametric and nonparametric survival analysis will be used to compare groups when time to recovery or time to relapse is of interest.

The power for hypothesis testing is sufficient with a the proposed sample size (n = 130). Simple correlations (r > 0.03), multiple regressions (see Project 1), and logistic regressions (OR > 2.67) all will have power > 0.80.

Back to the top

PROJECT 3: EVALUATING ALTERNATIVE APPROACHES TO MAKING THE DEFICIT VS. NON-DEFICIT DISTINCTION
Project Director: Michael Flaum, M.D.
Co-Investigators: Del Miller, Pharm.D., M.D., Stephan Arndt, Ph.D., Peg Nopoulos, M.D., and Nancy C. Andreasen, M.D.,Ph.D.
Funding Status: This project is not funded, and is being conducted with existing MH-CRC core resources.
Proposed Duration: The project will be pursued as described below for one year; thereafter, the project will be extended to examine external validators.

Specific Aims
1. To compare alternative approaches to categorizing patients with schizophrenia as deficit vs. non-deficit in terms of concordance, reliability and validity.
2. To develop a case identification method using existing MH-CRC phenomenological data which optimizes the validity of the deficit vs. non-deficit dichotomy.
3. To explore the phenomenological and neurobiological correlates of the deficit syndrome within the MH-CRC data bases.

Background and Rationale
Over the past two decades there has been a marked resurgence of interest in the negative symptoms of schizophrenia, which had been seen as fundamental to the disorder by both Kraepelin and Bleuler. A vast literature has emerged demonstrating relationships between negative symptoms and a wide variety of phenomenological and neurobiological variables (e.g., poor premorbid functioning, generalized impairment on neuropsychological tests, and increased likelihood of structural and functional brain abnormalities. The perceived importance of negative symptoms as a core symptom dimension in schizophrenia was reflected in the DSM-IV criteria by increasing the diagnostic importance placed on negative symptoms as one of the defining characteristic features of the disorder.

One important aspect of this body of research that remains unresolved has to do with how investigators and clinicians deal with the "primary vs. secondary" issue. It is apparent that there are several common factors which may result in the objective appearance of negative symptoms among individuals with schizophrenia. Examples include affective flattening secondary to neuroleptic induced akinesia, social withdrawal as a consequence of prominent hallucinations or delusions, apathy resulting from chronic institutionalization, and anhedonia in the context of a depressive syndrome. Each of these may closely mimic or exacerbate the appearance of those negative symptoms which presumably reflect an underlying neural deficit. Carpenter and colleagues have argued that if this etiological heterogeneity of negative symptoms is ignored, the signal to noise ratio will be so reduced as to minimize the likelihood of identifying neural mechanisms underlying those negative symptoms that may be a direct reflection of brain dysfunction They and others have demonstrated a variety of differential relationships between broadly defined and narrowly defined ("primary" and enduring) negative symptoms. They have suggested that as a starting point, we refer to those negative symptoms presumed to be etiologically related to underlying brain dysfunction as "deficit" symptoms, and use the term "negative symptoms" in a broader, purely descriptive sense.

There is no disagreement that our ability to tease apart primary from secondary negative symptoms will enhance the likelihood of identifying neural mechanisms underlying those that are indeed primary or "deficit." However, it is the optimal strategies for doing so which remain somewhat unclear, with at least two broad alternative approaches in use. One is to have raters make this distinction during the time of assessment, based on their clinical judgment and taking into account all available information. The other approach involves purely descriptive ratings of negative symptoms (irrespective of presumed underlying cause), along with descriptive ratings of the common secondary causes (e.g., EPS, depression, etc.). A variety of statistical methods can then be employed to try to account for the secondary causes. The Maryland group has developed specific methodology for both approaches. For the direct approach, they have developed an instrument (the Schedule for Deficit Syndrome (SDS)) which aids skilled clinicians in making the primary vs. secondary distinction based on clinical judgment. They and others have demonstrated good reliability for making the deficit vs. non-deficit distinction with it. The Maryland group has also developed a proxy case identification method which uses variables from the BPRS. With it, they found relatively high levels of sensitivity and specificity compared to ratings with the SDS, as well as significant predictive validity. However, they cautioned that although this method might be an acceptable proxy for large scale epidemiological samples, it "should be avoided in small-sample clinicopathologic correlation studies where classification of each subject is crucial."

The data bases collected thus far by the MH-CRC are somewhere in between; i.e., while they are not large scale epidemiological studies, they certainly go beyond "small samples." We are interested in optimally valid classification of subjects, but thus far we have not employed a direct method for making the deficit/non-deficit dichotomy (e.g., the SDS). The primary instrument we have used to evaluate negative symptoms has been the SANS. This was developed as a descriptive instrument in which raters are specifically instructed to "rate what they see, rather than what they think," in the effort to enhance reliability. It must be kept in mind that one of the reasons negative symptoms had been relatively de-emphasized in earlier diagnostic criteria (e.g., RDC, DSM-III and III-R) had to do with doubts about whether they could be assessed in a reliable and valid manner. The use of rating scales has allowed for their assessment with acceptable levels of reliability, although they still prove more difficult to rate than psychotic symptoms. Our concern has been that a requirement of further decision-making on the part of the rater regarding the etiology of negative symptoms may result in a substantial decrement in their reliability. Thus, our intended strategy has been to pursue a descriptive approach with the expectation of combining the negative symptom ratings with objective ratings of those factors which may cause secondary negative symptoms, (e.g., EPS scales, depression inventories, SAPS ratings, etc.). These factors can then be accounted for to some extent in multivariate analyses. This approach may be somewhere in between the BPRS-based proxy method, and that based on sophisticated clinical judgment in terms of validity.

We are interested in comparing approaches to making this distinction for several reasons. One is that the research on the deficit vs. non-deficit dichotomy has been increasingly generative and compelling in demonstrating a variety of phenomenological and neurobiological correlates which are potentially informative at the level of etiology (e.g., (Bustillo et al 1997; Waltrip et al 1997; Ross et al in press). Most of this work however, has come from a relatively small number of investigators who are addressing the issue directly (either with the SDS or the proxy approach proposed by the Maryland group (Kirkpatrick et al 1993)). Approaches which rely on statistical methods to control for secondary causes have been less generative, now more than a decade since the publication of Carpenter's heuristic presentation of the construct (Carpenter et al 1987). A second and related reason for our interest at this time is that we are in the process of developing a new assessment instrument which may largely take the place of the SANS and SAPS (see project 6 on the development of the "5-STEPS" in this section). As a part of this effort, we are trying to determine the optimal manner in which to incorporate information that will allow for the deficit/non-deficit distinction to be made in an optimally reliable and valid manner.

In this project, we seek to evaluate alternative approaches to the classification of deficit vs. non-deficit in terms of concordance of classification and potential validators. The ultimate aim is to employ the best method we can to differentiate deficit vs. non-deficit patients, and to exploit the MH-CRC phenomenological and neurobiological data bases to further examine the validity and correlates of the deficit/non-deficit dichotomy.
Sample hypotheses
1. The inter-reliability of the deficit vs. non-deficit syndrome as made via the SDS will be good, as indicated by Kappa values of > 0.60.
2. Inter-rater reliablity of the deficit vs. non-deficit syndrome via an unstructured interview will be lower than that achieved via the SDS.
3. Concordance between SDS classification and classification via an unstructured interview will be high, as evidenced by Kappa values of > 0.60.
4. Case classification algorithms based on multiple instruments from the MH-CRC data base will yield high levels of concordance with "gold standard" classification of deficit vs. non-deficit, as evidenced by Kappa values > 0.60.
5. Among patients with schizophrenia who are followed prospectively, those categorized as having the deficit syndrome by a case classification algorithm will be found to have more stable and persistent negative symptoms than those categorized as having secondary negative symptoms, both during 3 week medication free periods, and during a one year follow-up.

Experimental Design
Implementation of the SDS: The SDS will be administered to a total of ~100 subjects with schizophrenia over a one year period. These will include consecutive admissions to the MH-CRC research unit ( N ~ 40) as well as subjects in our prospective longitudinal study (Project 1 of this section) who are returning for 2, 5 or 9 year follow-up evaluations (N ~ 60), and who have consensus DSM-IV diagnoses of schizophrenia. This instrument will be incorporated into our weekly instrument training program (described in Section 5 of the Assessment and Training Core Unit), which is regularly attended by the research nurses, research assistants and MD fellows. For those patients on the inpatient unit, the instrument will be completed by their assigned primary research nurse, and independently by the MD fellow currently staffing the inpatient unit. Inter-rater reliability will be assessed with those 40 pairs of ratings. For those in the longitudinal follow-up study, it will be completed only by their assigned primary research assistant.

Classification based on unstructured interview: All those who participate in the diagnostic consensus conferences for both inpatients and longitudinal follow-up subjects (as described in Sections 4 of the Assessment and Training Core), will be familiarized with the deficit concept and the operationalized criteria for the deficit syndrome proposed by Carpenter et al (1988), through readings and discussions during a pilot phase. During the diagnostic consensus conferences we will follow these criteria in order to classify the same ~100 subjects as deficit or non-deficit. The SDS will have been completed prior to these presentations, and those who did them will not participate in this aspect of the study. Each of the items delineated in those criteria (i.e., 1) two of six negative symptoms; 2) enduring for > 1 year; 3) primary) will be added to our diagnostic consensus instrument. As we do now for our consensus diagnosis forms, raters will code their responses independently after the case is presented and the patient interviewed, and prior to any group discussion. These ratings will be entered in a data base, providing estimates of inter-rater reliability for each of the items and for the resulting deficit/non-deficit classification, in the absence of a structured interview. Following these independent ratings, a consensus will be established by the group regarding the deficit vs. non-deficit classification.

Generation of Gold Standard Categorization: Those subjects who are similarly classified by both methods (SDS and consensus) as either deficit or non-deficit and for whom both raters on the SDS agreed (in the reliability cases) will make up the "gold standard" group, which will serve as the primary validator of competing case identification algorithms (see below). Based on previous studies, we anticipate ~25 % to be classified as deficit. We are anticipating >75% agreement, yielding ~15 - 20 subjects with a concordant classification of deficit and 55 - 60 of non-deficit.

Generation of proxy case identification algorithms: A series of computer algorithms using variables from the battery of weekly ratings, the intake CASH and the intake PSYCH will be written and compared in terms of their concordance and predictive validity in dichotomizing patients with schizophrenia as deficit vs. non-deficit. These algorithms will consist of ratings of various combinations of items from instruments that assess positive and negative symptoms (SANS, SAPS, BPRS), depressive symptoms (Hamilton and Calgary scales), and various aspects of EPS (AIMS, Simpson Angus), along with longitudinal ratings of these symptoms (from the PSYCH and CASH). The specific combination of items, and variations in severity thresholds will be compared in an iterative manner in terms of their concordance with each other, with the BPRS-based case identification of Kirkpatrick et al (1993) and with the "gold standard" groups as defined above.

Evaluation of comparative validity of case identification algorithms: The comparative validity of the case identification algorithms will evaluated by three approaches: 1) concordance, sensitivity and specificity relative to "gold standard" classification; 2) strength of prediction regarding degree of stability of negative symptoms throughout a three week medication washout period; and 3) strength of prediction regarding degree of stability of negative symptoms over a one year follow-up period. The first will involve straightforward calculations of the concordance, sensitivity and specificity of different case classification algorithms with that of the gold standard classification. This will be limited to those subjects who have concordant SDS and consensus classification (N ~ 75). Both of the other approaches will take advantage of existing data on larger numbers of subjects. For approach #2, this will include all those who have participated in the three week medication wash-out protocol within the MH-CRC (~ 250 subjects with schizophrenia). Here, we will compare classification of deficit vs. non-deficit by each algorithm in terms of the degree to which SANS global ratings are sensitive to changes in medication status. At the simplest level, the validity of the algorithm is increasingly supported by the degree to which non-deficit group evidenced a greater change in SANS ratings than the deficit group when medications were discontinued. For approach #3, we will use existing data gathered from the longitudinal study described above in project 1 to evaluate the comparative validity of the case classification algorithms. In that study, weekly ratings are made of each global negative symptom of the SANS. We will examine the intra-individual variability in these ratings over a one year time period, with the prediction that the non-deficit group will have significantly greater variability on these ratings than the deficit group. We have one year follow-up data on ~200 subjects with schizophrenia, for inclusion in this analysis.

Data Analysis: Reliability (Hypotheses 1 and 2) will be quantified by Kappa for the deficit vs. non-deficit classification. Separate reliability coefficients will be generated for agreement between two or more expert raters (e.g., Nancy C. Andreasen, Michael Flaum, Del Miller, Peg Nopoulos) vs. reliability between non-experts, (e.g., CRC-related trainees). Hypothesis 3 will also be quantified by Kappa for the deficit vs. non-deficit categorization according to the SDS vs. consensus unstructured interview approach. Hypothesis 4 will be tested by a series of general linear models predicting changes on and off medication based on the deficit vs. secondary negative distinction (according to different algorithms), changes in extrapyramidal symptoms, changes in positive symptoms, and their interactions. Hypothesis 5 will be tested by quantifying the standard deviation of the mean weekly ratings for each global negative symptom over a one year time period. Those patients classified as having secondary negative symptoms would be expected to have significantly greater variance than the secondary negative group.

Future Directions: Once an optimal method for identifying a deficit subgroup is determined by the methods described above, we envision a series of studies in which subjects so classified would be compared to non-deficit schizophrenics, controls, or other comparison groups on a wide variety of phenomenological and neurobiological measures. These would include those variables that we and others have shown to be specifically related to negative symptoms (e.g., poor premorbid functioning, greater likelihood of neurological soft signs, generalized cognitive deficits, decreased hippocampal volume, decreased frontal, pontine and cerebellar blood flow) as well as those variables within the MH-CRC data base that have been demonstrated by others to be differentially related to the deficit vs. non-deficit groups (e.g., season of birth, depressive symptoms, specific neuropsychological deficits, etc.).

Back to top

PROJECT 4: EVALUATION OF PREMORBID COGNITIVE FUNCTION USING CHILDHOOD SCHOLASTIC APTITUDE TEST SCORES
Principal Investigator: Peg Nopoulos, M.D.
Co-Investigators: Michael Flaum, M.D., Del Miller, Pharm.D., M.D., Stephan Arndt, Ph.D., Nancy C. Andreasen, M.D., Ph.D.
Funding Status: Funded by an Iowa Mental Health Grant to Dr. Nopoulos. Application for NIH support is planned after adequate preliminary data are obtained.
Proposed Duration: This project has been ongoing for one year and will continue for at least another two years.

Specific Aims
1. To retrospectively obtain childhood standardized cognitive test scores using a unique and superior quality data base in order to assess premorbid cognitive function in a group of adult patients with schizophrenia.
2. To assess whether schizophrenia is associated with premorbid cognitive deficits, and if so, the longitudinal, but still premorbid, course of the deficits.
3. To evaluate the relationship of phenomenological variables such as subtype of schizophrenia, symptoms, course of illness and treatment response to premorbid cognitive functioning.
4. To determine if other measures of developmental abnormality (e.g., poor premorbid social adjustment, developmental brain anomalies, etc.) are associated with premorbid cognitive dysfunction.

Background and Rationale
The Department of Education at the University of Iowa possesses a superior quality data base of educational records unparalleled by any other state or institution in the nation. Beginning in the early 1930's, the Iowa Testing Program was formed and this institution has been responsible for the development, standardization and distribution of the cognitive tests administered to school aged children in Iowa for more than 50 years and to the rest of the nation for over 30 years (Peterson 1983). The Iowa Test of Basic Skills and the Cognitive Abilities Test are the two most widely used. Scores of these two tests from Iowa school children have been meticulously archived and maintained by the Department of Education. Coupled with the uniqueness of this data base is the geographic stability of the Iowa population, affording the opportunity to study Iowans longitudinally. This scenario creates a superior opportunity to retrospectively assess the cognitive function of patients with schizophrenia prior to the onset of their illness.

There is now a large literature documenting the presence of significant cognitive impairment in patients who suffer from schizophrenia. However, despite the large amount of research on the association between cognitive function and schizophrenia, there is currently no comprehensive theory regarding the nature of the relationship. It is not yet known whether cognitive function deteriorates as a result of the illness, whether individuals with poor cognitive function are more susceptible to schizophrenia, or whether cognitive function is related to phenomenologic variables such as course of the illness or symptomatology.

Identifying the course of cognitive function in patients with schizophrenia may well provide clues to the nature of the association with the illness. Several recent studies have shown that cognitive dysfunction is present very early on in the illness and may well be stable over time (Hoff et al 1992; Nopoulos et al 1994; Sweeney et al 1991). Therefore, if the deficit is present when patients present with psychosis, when does the deficit begin? Does it occur concomitantly with the overt expression of the illness or does it pre-date the onset? A few prospective studies evaluating premorbid IQ are limited by following only those patients who are at genetic high risk for the illness (offspring of schizophrenic mothers). As most patients with the illness do not have such heavy genetic loading, this restricts the sample considerably. Retrospective methods, however, encompass a larger, more etiologically heterogeneous sample. There has been relatively few retrospective studies evaluating premorbid IQ in patients with schizophrenia, most of them published 20-30 years ago. In a review by Aylward, only four retrospective studies evaluating premorbid IQ were identified for meta-analysis (Aylward et al 1984). Results of the meta-analysis indicate that there is, indeed a difference in IQ between preschizophrenic patients and their peers, the preschizophrenic patients scoring much lower. Furthermore, these findings were primarily a male effect.

Postmorbid cognitive deficits in schizophrenia have shown association with a variety of phenomenological measures such as poor response to treatment subtype, and severity of negative symptoms. However, there is an inherent methodological flaw in evaluating postmorbid cognitive dysfunction and phenomenological variables - one cannot separate out the effect of the illness itself on cognitive function. Evaluating the relationship between premorbid cognitive function and postmorbid phenomenology would eliminate this confounder.

Hypotheses:
1. The schizophrenic group as a whole will demonstrate premorbid cognitive dysfunction by having test scores consistently in the lower 50th percentile rank compared to state and national norms.
2. Low premorbid cognitive function will be associated with male gender, early age of onset, predominance of negative symptoms, poor response to treatment, non-paranoid subtype, and poor outcome.
3. Low premorbid cognitive function will be associated with a variety of structural brain abnormalities that are considered to be neurodevelopmental in origin: decrement in total brain tissue, increase in total CSF, gray matter heterotopia, etc.

Methods:
Patient Sample: A sample of 100 patients with schizophrenia will be identified. Next, consent to release information of standardized test scores will be obtained from the patients. Some of this may be obtained retrospectively - by contacting the patient by phone, outlining our request, and mailing them a consent to release information form for them to sign and return. This process is tedious and can be done effectively on a minority of the sample. However, the 260 patients enrolled in our longitudinal protocol (see Project 1) have regular contact with the MH-CRC (every 6 months). Therefore, we are obtaining release of information forms as part of routine follow-up in this sample. In addition, as patients come into the research center, we can ask them at intake to sign a release of information. To date, we have obtained consents from approximately 70 patients.

Data Analysis
Hypothesis 1 (there is a preponderance of scores below the 50th percentile) will be tested using a simple binomial test. This can also be verified using the actual percentile ranks tested against a predicted value of 50. Power for these tests are excellent. A mean percentile rank of 0.40 in the group represents a large effect size (Cohen's d = 0.35) so that a sample of 100 has a power > 0.93 of detecting this difference.

Hypotheses 2 and 3 will use simple regression methods to find an association between the percentile rank scores and current outcome and brain volume measures. Patient scores will also be correlated with phenomenologic and neurobiological measures. Examples of phenomenologic measures would include age of onset, severity of symptoms, types of symptoms, subtype of illness, response to treatment, course, and outcome. Examples of neurobiological measures includes brain structure as measured by MRI, obstetrical complications, and neurological soft signs. Even controlling for covariates such as parental SES, we expect these regression analyses to have power greater than 0.76 (n = 100, 4 covariates with R2 = 0.3, alpha = 0.05, and the premorbid r2-added = 0.05).

Back to top

PROJECT 5: CHARACTERIZATION OF CHRONIC SCHIZOPHRENIA IN LATE LIFE: NEUROPSYCHOLOGICAL AND SYMPTOM ASSESSMENT OF ELDERLY SUBJECTS
Principal Investigators: Susan K. Schultz, M.D. and Jane S. Paulsen, Ph.D.
Co-Investigators: Peg C. Nopoulos, M.D., Del D. Miller, Pharm.D., M.D., Daniel S. O'Leary, Ph.D., Michael Flaum, M.D., Stephan Arndt, Ph.D.
Funding Status: Funded by an Iowa Mental Health Grant to Dr. Schultz. Application for NIH support is planned after adequate preliminary data are obtained.
Proposed Duration: This project has been going on for two years and will continue for at least another two years.

Specific Aims
1. To examine a series of elderly nursing home subjects with and without schizophrenia, comparing measures of cognitive impairment as well as measures of positive and negative symptoms.
2. To test the hypotheses that elderly patients with schizophrenia who develop cognitive impairment in late life have greater severity of negative symptoms compared to patients with schizophrenia without cognitive impairment.
3. To examine the relationship between negative symptoms and cognitive impairment in schizophrenic patients compared to non-schizophrenic control subjects matched for age and demographic features. This will help determine whether negative symptoms may be associated with age-related cognitive decline irrespective of the diagnosis of schizophrenia.

Background and Rationale
Previous research has suggested that the neuropsychological profile of schizophrenia involves impaired executive functioning (Abbruzzese et al 1993; Cummings 1993; Freeman et al 1993; Pantelis et al 1992; Paulsen et al 1994; Sullivan et al 1994) as well as impaired motor and memory function (Delis et al 1987; Matthews et al 1964). However, there is little consensus as to whether the effects of aging magnify these impairments, ameliorate them, or perhaps impart a predisposition to overt dementia. Interestingly, the literature is quite variable and not infrequently reports an improvement in neuropsychological measures over the life course of the illness (Heaton et al 1987). In recent years there has been interest among geriatric researchers in the cognitive features of late-onset schizophrenia (Jeste 1993), yet there is surprisingly little consensus about the long-term neuropsychological outcome of the more common early-onset illness in patients who are now approaching late life, particularly in terms of the impact of symptom severity and chronic treatment (Strauss 1993).

In non-geriatric patients with schizophrenia, negative symptoms are associated with many abnormalities, including poor visuospatial processing and poor motor speed (Benton 1974b; Strauss 1993), implicating disruptions of the complex interactions within frontal-striatal and cerebellar systems. Whether these impairments manifest to an even greater extent in the context of aging and whether they are directly related to negative symptom severity represents an important area for future work. Unfortunately, the population of elderly individuals with typical early-onset schizophrenia is not one that will readily present itself to the researcher within a tertiary setting. Elderly schizophrenic patients with negative symptoms may be severely symptomatic but may not constitute a behavioral problem and consequently are rarely if ever referred for an acute evaluation. Thus this group remains largely "invisible" to researchers and clinicians yet is a source of significant morbidity in the elderly population. Investigations of this group require a concerted and systematic effort on the part of the researcher involving the cooperation of not only the patients but the caregivers, conservators and administrators within the facilities providing care. As negative symptoms uniquely appear to persist or increase with aging (in contrast to positive symptoms) (Schultz et al 1997a) their importance in studies of the elderly has become increasingly clinically relevant, particularly as there are now novel antipsychotic medications targeting negative symptoms as well as new medications for early cognitive dysfunction (e.g., risperidone and donezepil, respectively).

It is important to consider that negative symptoms may not be unique to schizophrenia but rather may be a consequence of neuropsychological decline in the non-psychiatrically ill elderly. Negative symptoms have been a focus of study in schizophrenia but may also appear secondary to a number of other conditions. For example, negative symptoms secondary to Alzheimer's disease have recently been an area of investigation (Reichman et al 1996). Negative symptoms may be categorized into four groups: 1) alogia - a decreased quantity or content of speech, 2) affective flattening - a decrease in emotional responsiveness in speech, gestures and expressions, 3) avolition/apathy - a loss of persistence in volitional activity and 4) anhedonia/asociality - a loss of interest in relationships and activities. Any or all of these symptoms could result from a decline in cognitive function. This creates a "chicken or the egg" dilemma in which patients with schizophrenia and negative symptoms may be at risk for neuropsychological dysfunction - or the converse may be true, i.e., patients with schizophrenia and neuropsychological decline may develop more negative symptoms. To address this question, this project will assess a group of non-schizophrenic control subjects which may help to tease apart this relationship in the absence of schizophrenia.

This project seeks to begin systematic assessment of geriatric populations of subjects in various Iowa facilities (Knoxville VA Nursing Home Care Unit, Marshalltown Veterans Home, and the Independence Mental Health Institute) with the following goal: To characterize the association between cognitive impairment and negative symptom severity in nursing home residents by performing on-site assessments including the Mattis Dementia Rating Scale (Mattis 1976), Scales for the Assessment of Positive and Negative Symptoms (SAN/SAPS) a neuropsychological battery and the Comprehensive Assessment of Symptoms and History (CASH) (Andreasen et al 1992a). We hypothesize that negative symptoms in late life will correlate with severity of cognitive impairment in both the group with schizophrenia and the healthy comparison group, but the patient group will have greater severity of negative symptoms relative to degree of cognitive impairment.

Methods
Subjects:
Inclusion Criteria: Subjects must have a primary diagnosis of schizophrenia-spectrum disorder to include schizophrenia, schizoaffective disorder or delusional disorder, and currently reside in a nursing home facility. Comparison subjects must not have schizophrenia, though those with a history of other psychiatric diagnoses or alcohol abuse in remission will be included if there is no evidence of an active problem. Subjects must be over the age of 60 years.

Exclusion Criteria: Subjects with mental retardation, severe head trauma, overt dementia of the multi-infarct or Alzheimer's type will be excluded. Serious medical illness impacting on cognition will be excluded. Patients with serious medical illness accounting for delirium or a dementia syndrome will be excluded, however, patients with medical illness may be recruited provided the medical condition does not contribute to a delirious state or prevent cooperation with testing (as in a severe aphasia). Most importantly, patients with depressive syndromes will be excluded as these symptoms overlap too extensively with negative symptoms. It is necessary to exclude patients with active major depressive disorder, as this condition will overlap with ratings of negative symptoms, however we will plan to include patients with a history of depression provided they are not experiencing an active state of depression. Through this flexibility in recruitment criteria i.e., including medical illness and subjects with previous depression, our ability to obtain an adequate sample is likely to be enhanced.

Subjects will be assessed at their respective facilities by a research assistant and/or a postdoctoral fellow. A screening and review of records on-site by the postdoctoral fellow will be performed to identify appropriate subjects who are able to cooperate with further assessment. Subjects will then undergo a diagnostic evaluation in a consensus evaluation with the MH-CRC psychiatry and neuropsychology team including a review of all available records and a review of a videotaped interview. Within the Knoxville Nursing Home Care Unit and the Iowa Veterans Home, a sample a nursing home residents without psychiatric illness will be recruited with matching demographic features to obtain a comparison group.

In conjunction with the above screening process, the postdoctoral fellow will administer a Mattis Dementia Rating Scale and SANS/SAPS ratings to generate a broad data base to include patients who may not be able to cooperate with a more intensive neuropsychological assessment. The postdoctoral fellows and research assistants will be trained to perform this battery in collaboration with the MH-CRC Cognitive Neuroscience Core Unit and under the consulting guidance of Dr. Jane Paulsen, neuropsychologist in the Departments of Psychiatry and Neurology. It is estimated that approximately 50 patients will be able to complete the full neuropsychological assessment over the next year, as we have already identified at least this many through preliminary screening of the various facilities. It is further expected that approximately 50 additional subjects may be assessed with the Mattis and SAN/SAPS on-site over the next year.

Assessment:
Diagnostic Assessment:
Diagnostic assessment with consensus diagnosis reached by interview and record review.
Symptom Rating Scales for Current Symptomatology:
SANS and SAPS, HAM-D (Hamilton Depression Rating Scale)
Cognitive Screening:
Mini-Mental Status Exam, Mattis Dementia Rating Scale*
*The Mattis was selected because it covers a broad range of cognitive domains and is created specifically to identify impairments in the elderly. It has a total score of 144 with subdomains targeting specific areas of impairment, and can be administered in approximately 1/2 hour.
Neuropsychological Assessment:
Boston Naming Test, Circle A Test, California Verbal Learning Test, Logical Verbal Memory, Benton Visual Retention Test, Trails A and B Test, Finger Tapping Test, WAIS, Stroop Test, ANART (provides a measure of premorbid intellectual function), and Wisconsin Card Sort.

Data Analysis
At least two analyses will be completed. The first of these analyses will examine whether cognitive decline is more frequently associated with negative symptoms rather than positive psychotic or disorganized symptoms in subjects with schizophrenia. This will involve the screening information obtained on-site including the Mattis Dementia Scale and SAN/SAPS. Impairment on the Mattis will be employed as a continuous variable, as it is likely that all degrees of impairment will be seen on a continuum as opposed to a discrete dichotomy of presence/absence of cognitive impairment. The Mattis scores will be employed as dependent measures, while the SANS/SAPS scores will be used to categorize the schizophrenic patients into three groups (dimensions) in terms of psychotic, negative and disorganized symptom severity, depending on which cluster of symptoms predominates. These three dimensions will be used as class variables in an analysis of variance to determine the association of each with cognitive impairment. For subjects who are able to cooperate with more detailed assessments, specific variables from tests of executive functioning, for example, will be assessed as dependent measures. The second type of analysis will examine the correlation between cognitive impairment and negative symptoms in patients with schizophrenia compared to subjects without schizophrenia, as described below in the power analysis.

Estimated sample needed for projected effect size: The Knoxville VA Medical Center, as an example site, has approximately 260 nursing home residents with approximately 100 patients diagnosed with chronic schizophrenia. Conservatively estimating that only one-half will consent to participate in the minimum assessment of a Mattis Dementia Rating Scale and a SAN/SAPS rating, then 50 subjects with schizophrenia and 50 age-matched patients without schizophrenia, but with a chronic affective illness, chronic anxiety disorder or other non-psychotic chronic illness will be compared as control subjects. Using a correlation coefficient across the sample to examine the association of negative symptoms and cognitive impairment, we hypothesize that negative symptoms are negatively correlated with cognitive impairment as measured by the Mattis Dementia Scale, i.e., a higher SANS score would be associated with a lower Mattis score. With an at 0.05 and power of 80 we estimate smallest correlation equivalent we would like to detect is 0.278. However, we expect to see a larger correlation (e.g., r = 0.50) in the schizophrenia group and a near zero correlation within the non-schizophrenia patients. We will use a two-tailed test (with Fisher's r-to-Z transform) to contrast these two independent correlation coefficients. The power to detect this difference is greater than 0.75 with 50 subjects in each group.

Back to top

PROJECT 6: DEVELOPMENT OF A NEW SCALE TO MEASURE PSYCHOPATHOLOGY IN SCHIZOPHRENIA: THE "5 STEPS"
Principal Investigator: Michael Flaum, M.D. (Iowa)
Co-Investigators: Nina Schooler, Ph.D. (LIJ-Hillside), Steve Marder, M.D. (UCLA), Robert Buchanan, M.D. (Maryland)
Funding Status: Funding for this project has been applied for through a NARSAD Established Investigator Award (Dr. Schooler) 10/97; funding decision currently pending.
Proposed Duration: This project has been underway for two years. The ongoing work proposed herein is expected to begin in the Spring of '98 and continue for 18-24 months thereafter.

Specific Aims:
1. To perform multi-center reliability studies of the current version of the "5-STEPS" instrument.
2. To establish concurrent validity by comparing 5-STEPS ratings with independent ratings of existing instruments, including the BPRS, SANS/SAPS, and PANSS.
3. To evaluate sensitivity to change in this instrument vs. existing instruments.
4. To make revisions to the current version of the instrument as guided by these reliability and validity studies.

Background and Rationale
As discussed in the Assessment and Training Core Unit, our group has been collaborating with several other schizophrenia researchers over the past few years in developing a new instrument, primarily designed to assess change in psychopathology during treatment trials in schizophrenia. This effort was initiated by Nina Schooler, and has been guided by her, with limited financial support from the pharmaceutical industry. The impetus for the effort stemmed primarily from dissatisfaction with existing instruments, and from ongoing controversy within the field as to the most appropriate instrument(s) to use in treatment trials of schizophrenia. Typically, the decision comes down to one or more of the following instruments: The BPRS, the SANS, the SAPS or the PANSS (Overall, 1962; Andreasen, 1983; 1984; Kay, 1988). Each has advantages and disadvantages. The BPRS has been in use the longest, and largely by virtue of that is considered the "industry standard." However, it was not designed specifically for use in schizophrenia, and has inadequate coverage of many core symptoms, particularly negative and disorganized symptoms. The PANSS expands upon the BPRS with respect to negative symptoms, but otherwise retains its basic content. The SANS and SAPS provide much broader coverage of schizophrenic symptomatology. However, many designers of clinical trials have chosen against it, primarily because its length and complexity renders it relatively inefficient and impractical to administer and process in the course of brief evaluations.

Further, none of these instruments adequately reflect our current understanding of symptom dimensions in schizophrenia. Over the past decade, there have been dozens of studies examining this issue (reviewed in (Andreasen, 1995c) using a variety of instruments and analytic methods. While the specific number of factors or dimensions varies to some degree across studies, almost all identify at least three factors: negative symptoms, "psychotic symptoms" (delusions and hallucinations), and disorganized symptoms (disorganized speech, behavior and inappropriate affect). Many studies also identify additional factors of depression/anxiety and excitement as independent. A recent study, which employed our DSM-IV field trial data base suggested that a similar factor structure exists in non-schizophrenic psychotic disorders (Ratakonda et al 1998) complementing earlier studies which demonstrated a similar factor structure in schizotypy (Gruzelier 1995). This growing literature suggests that each of these symptom factors may correspond to a discrete pathophysiological condition which may be differentially responsive to treatment interventions. Acquisition and analysis of symptom data organized according to these dimensions may therefore increase both the sensitivity and specificity of detecting efficacy.

With these considerations in mind, Dr. Schooler convened a large group (~25 individuals) of clinical and industry investigators in 1995, with the goal of developing an instrument which would incorporate contemporary understanding of the psychopathology of schizophrenia, and which might become the new "industry standard." Included in this group were the developers of the major existing instruments, including Dr. Overall (BPRS), Dr. Andreasen (SANS/SAPS) and Dr. Lindenmeyer (PANSS). Also included were representatives from four of the five currently funded schizophrenia-oriented MH-CRC's (Iowa, UCLA, Maryland and LIJ-Hillside).

Progress Report
The initial effort involved identifying the limitations of the existing instruments and determining the need of the field. The clear consensus was that it was an appropriate time to develop a new instrument, largely for the reasons described above. The next stage, conducted over the following several months, involved surveying all existing instruments (including diagnostic instruments such as the SADS, CASH, DSM-IV field Trial instrument and the SCAN) and prioritizing items and constructs in terms of their clarity (user-friendliness), specificity (vs. redundancy or overlap with other items), reliability and coverage of core dimensions of schizophrenic psychopathology, as informed both by expert opinion and the existing literature. Our group took the lead in drafting the resulting items into an initial version of a new instrument, which was then distributed to the group as a whole and modified according to feedback. Several items which had not been included on any existing scales were added, with the idea that over-inclusiveness at this early stage was preferable to omissions. A second meeting was convened in which the resulting instrument was presented and discussed in detail, and further modified accordingly. Our group then piloted the resulting version of the instrument in the context of our weekly consensus diagnostic conference (described in the Assessment and Training Core Unit) with 15 patients. Drs. Andreasen or Flaum conducted the interviews, and ratings on the new instrument were made by between 5 - 12 clinical researchers for each patient. The data were then analyzed for item reliability (intraclass r), base rates (mean and SD) and internal consistency among items (Cronbach's alpha).

For the reliability analyses, the raters were divided into "expert" and "non-expert" groups based on years of experience with the use of assessment instruments. Among the "expert" raters, the median intraclass r for the 5 negative symptom items was 0.61 (range 0.52 - 0.77) and for the 5 psychotic symptom items, it was 0.85 (range 0.80 -0.88). For the 5 disorganized symptoms, the mean intraclass r among the experts was only 0.43 (range 0.0 - 0.68), but this must be interpreted in the context of very low base rates for each of these items (mean of < 1 for each disorganized item on a 0 - 5 scale, as opposed to > 1 for each of the psychotic and negative symptoms), yielding the reliability estimates unstable for those items. The version of the instrument used at the time categorized all of the remaining 10 items into an "other symptom" group, and the reliability was quite variable for these items, with a mean intraclass r of only 0.42 (range 0.15 - 0.79) among the experts. As expected, the "non-expert" rater group generally had lower inter-rater reliabilites: (mean intraclass r's of 0.43, 0.67 and 0.44 and 0.43 for the four categories of symptoms respectively). In light of the fact that these reliability data were gathered with a version of the instrument that did not yet have anchors, and that no specific training had been provided, the overall impression of this pilot study was encouraging. Specifically, the reliability estimates for both the negative and psychotic symptoms were comparable to those demonstrated by multiple prior studies in this and other labs using the SANS/SAPS, the CASH, and the DSM-IV field trail instrument. This was particularly important to demonstrate because of concerns that reliability would be markedly compromised by the absence of subscale item ratings (i.e., in the absence of subscale items used in the SANS/SAPS such as "change in facial expression" to aid in the rating of "affective flattening").

Each of the 15 subjects included in this pilot study had been inpatients on the MH-CRC unit at the time, and therefore had the usual battery of weekly ratings completed as well, typically by one of the research nurses. We were therefore able to get a rough feel for validity by comparing the mean expert ratings on the new instrument (excluding those of the research nurse assigned to the patient) to the corresponding SANS/SAPS or BPRS ratings made during the same week (there was a corresponding item in one of the two scales for all but 3 of the items). The correlations between the new instrument ratings, and the SANS/SAPS and BPRS ratings were consistently high: The median r for negative symptoms was 0.70 (range .52 - .72); for 5 psychotic symptoms the median r was 0.82 (range .64 - .95); and for disorganized symptoms the median r was 0.74 (range .66 - .79). Finally the internal consistency of items within groups was high as evidenced by Chronbach's alpha coefficients of 0.92, 0.92 and 0.84 for negative, psychotic and disorganized symptoms respectively. As expected it was lower for the "other symptom" group (0.57) which included symptoms which were negatively inter-correlated with one another (e.g., pressured speech and fatigue).

These data were taken back to the group in a third meeting last year, after which the most recent round of modifications were made. In addition to modifications, eliminations and additions of some items, this revision also includes division of the "other symptom" category into "depressive/anxiety" and "activation" subscales. Finally, the Hillside group (Delbert Robinson and Meg Woerner) wrote a set of anchors and improved definitions for each of the items. The instrument was also formally named at that time: the 5 STEPS, an acronym for 5-dimensional Scale To Evaluate Psychopathology in Schizophrenia. The current version (4.0) is included in the Appendices #2, Interview Instruments. It consists of a total of 25 items (5 subscales of 5 items), each rated on an anchored 0 - 6 Likert scale. There was consensus among the group that development had evolved to a point that multi-center reliability and validity studies were now appropriate, as described below.

Hypotheses:
1. The overall and subscale reliability of the 5-STEPS will be good to excellent (e.g., projected median intraclass r > 0.70 both within and across sites, and > 0.60 for the corresponding test-retest design).
2. The overall and subscale concurrent validity will be good (e.g., projected median Pearson correlations of > 0.60 for the 5-STEPS with corresponding items from the SANS/SAPS, PANSS and BPRS).
3. Effect sizes of the 5-STEPS when administered on vs. off medications will be equivalent to or greater than those of the existing instruments.

Methods
Overview: Reliability and validity studies will be conducted at four MH-CRC sites: Iowa, LIJ-Hillside, Maryland, and UCLA. Two types of multicenter reliability studies will be conducted: one involving videotaped interviews to be completed across sites, and the other involving live patient interviews at each site. The latter will include both inter-rater and test-retest reliability designs (analogous to those we employed in the DSM-IV Field Trial Study (Flaum et al in press) Validity studies will entail comparisons between serial ratings made with the 5-STEPS to those of other standard instruments including the BPRS, SANS/SAPS and PANSS in patients for whom treatment status is changing.

Raters: Each site will provide a minimum of 4 raters, selected from their respective assessment core units. For the purposes of this phase of the study, it is expected that raters will have had substantial experience in the use of rating scales in this population, although their training and backgrounds may vary. All raters will complete a rater profile form (used in the DSM-IV Field Trial) to quantify their demographics, background and experience.

Patient subjects: Subjects will be drawn from those participating in ongoing protocols within each of the four MH-CRC's, and whose DSM-IV diagnosis is one of those included in the "schizophrenia and other psychotic disorders" section of that manual. At least half of the patients at each site must be in a protocol that requires a clear change in treatment over a period of 4 weeks, with a reasonable expectation of a resulting change in clinical status. This would include patients who are participating in a medication wash-out protocol, and those participating in various treatment protocols that include an off-medication lead in phase.

Across-site inter-rater reliability studies: Each site will generate 5 videotaped interviews from which ratings on the 5-STEPS can be made. Brief narrative summaries will accompany each videotape to provide basic background information and to fill gaps in the interview necessary for making ratings. (Similar to what is done for our SANS/SAPS reliability and training program). All 20 videotapes will be rated at all sites by a minimum of 4 raters / site.

Within-site reliability studies: Ratings will be made with the 5-STEPS on 20 patients at each site by 2 raters concurrently (one serving as the primary interviewer and the other observing and allowed at ask clarifying questions). The same patient will be interviewed by another pair of independent raters the following day. All ratings are to be based on the "worst over the previous one week," and all raters will have access to the current medical record.

Comparison with other instruments: As each of the MH-CRC's participating in this project collects the BPRS and either the SANS/SAPS or the PANSS routinely on their research subjects, these ratings will be used for comparison. It will be specified that: 1) these ratings will be completed by someone other than those making the 5-STEPS ratings; and 2) ratings of these instruments be made within 72 hours of both sets of 5-STEPS ratings.

Evaluation of sensitivity to change: For at least half of the subjects at each site, the 5 STEPS, BPRS, SANS/SAPS or PANSS ratings will be repeated between 3-6 weeks following initial evaluation in the context of a change in treatment status.

Data Analysis
Hypothesis 1 will be assessed using intraclass correlations. Power for detecting a significant reliability for the proposed N of 80 subjects is > 0.99. Of more importance is the confidence interval around our estimates of reliability. Given an N of 80 and a projected intraclass r of 0.70 (for the inter-rater design), the 95% confidence interval is 0.57 - 0.80. Inter-site variability in reliability will be assessed using Fisher's r to Z transforms and chi square tests (as per Hays 1973). Pearson correlations will be used to assess concurrent validity (Hypothesis 2). As in the case of Hypothesis 1, the power to detect significant correlations >0.99, with similarly sized confidence intervals. For Hypothesis 3, effect sizes will be calculated for each instrument with respect to symptom changes as a function of change in medication status.

© The University of Iowa 2005. All rights reserved.

Latest update May 7, 2005 Webpages maintained by Hans J. Johnson. E-mail the webmaster