Meta-analysis of binary data: which within study variance estimate to use?

We applied a mixed effects model to investigate between- and within-study variation in improvement rates of 180 schizophrenia outcome studies. The between-study variation was explained by the fixed study characteristics and an additional random study effect. Both rate difference and logit models were used. For a binary proportion outcome p(i) with sample size n(i) in the ith study, (circumflexp(i)(1-circumflexp(i))n)(-1) is the usual estimate of the within-study variance sigma(i)(2) in the logit model, where circumflexpi) is the sample mean of the binary outcome for subjects in study i. This estimate can be highly correlated with logit(circumflexp(i)). We used (macronp(i)(1-macronp)n(i))(-1) as an alternative estimate of sigma(i)(2), where macronp is the weighted mean of circumflexp(i)'s. We estimated regression coefficients (beta) of the fixed effects and the variance (tau(2)) of the random study effect using a quasi-likelihood estimating equations approach. Using the schizophrenia meta-analysis data, we demonstrated how the choice of the estimate of sigma(2)(i) affects the resulting estimates of beta and tau(2). We also conducted a simulation study to evaluate the performance of the two estimates of sigma(2)(i) in different conditions, where the conditions vary by number of studies and study size. Using the schizophrenia meta-analysis data, the estimates of beta and tau(2) were quite different when different estimates of sigma(2)(i) were used in the logit model. The simulation study showed that the estimates of beta and tau(2) were less biased, and the 95 per cent CI coverage was closer to 95 per cent when the estimate of sigma(2)(i) was (macronp(1-macronp)n(i))(-1) rather than (circumflexp(i)(1-circumflexp)n(i))(-1). Finally, we showed that a simple regression analysis is not appropriate unless tau(2) is much larger than sigma(2)(i), or a robust variance is used.

The Yugoslavia Prospective Lead Study: contributions of prenatal and postnatal lead exposure to early intelligence.

To investigate associations between the timing of lead (Pb) exposure on early intelligence, we examined the results of psychometric evaluations at ages 3, 4, 5, and 7 years, from 442 children whose mothers were recruited during pregnancy from a smelter town and a non-lead-exposed town in Yugoslavia. We compared the relative contribution of prenatal blood lead (BPb) with that of relative increases in BPb in either the early (0-2 years) or the later (from 2 years on) postnatal period to child intelligence measured longitudinally at ages 3 and 4 (McCarthy GCI), 5 (Wechsler Preschool and Primary Scale of Intelligence-Revised, WPPSI-R IQ), and 7 (Wechsler Intelligence Scale for Children-version III, WISC-III IQ), controlling for: Home Observation for Measurement of the Environment (HOME) quality; maternal age, intelligence, education, and ethnicity; and birthweight and gender. Elevations in both prenatal and postnatal BPb were associated with small decrements in young children's intelligence.

Suicidal behavior in bipolar mood disorder: clinical characteristics of attempters and nonattempters.

OBJECTIVE: Bipolar Disorder is associated with a higher frequency of attempted suicide than most other psychiatric disorders. The reasons are unknown. This study compared bipolar subjects with a history of a suicide attempt to those without such a history, assessing suicidal behavior qualitatively and quantitatively, and examining possible demographic, psychopathologic and familial risk factors. METHODS: Patients (ages 18 to 75) with a DSM III-R Bipolar Disorder (n = 44) diagnosis determined by a structured interview for Axis I disorders were enrolled. Acute psychopathology, hopelessness, protective factors, and traits of aggression and impulsivity were measured. The number, method and degree of medical damage was assessed for suicide attempts, life-time. RESULTS: Bipolar suicide attempters had more life-time episodes of major depression, and twice as many were in a current depressive or mixed episode, compared to bipolar nonattempters. Attempters reported more suicidal ideation immediately prior to admission, and fewer reasons for living even when the most recent suicide attempt preceded the index hospitalization by more than six months. Attempters had more lifetime aggression and were more likely to be male. However, attempters did not differ from nonattempters on lifetime impulsivity. LIMITATIONS: The generalizability of the results is limited because this is a study of inpatients with a history of suicide attempts. Patients with Bipolar I and NOS Disorders were pooled and a larger sample is needed to look at differences. We could not assess psychopathology immediately prior to the suicide attempt because, only half of the suicide attempters had made attempts in the six months prior to admission. Patients with current comorbid substance abuse were excluded. No suicide completers were studied. CONCLUSIONS: Bipolar subjects with a history of suicide attempt experience more episodes of depression, and react to them by having severe suicidal ideation. Their diathesis for acting on feelings of anger or suicidal ideation is suggested by a higher level of lifetime aggression and a pattern of repeated suicide attempts.

Modelling the decline pattern in functional measures from a prevalent cohort study.

In studying decline among cognitively impaired people, a prevalent cohort study design is commonly used to account for entry into the study at different levels of impairment. The data set typically consists of many short series of repeated measurements collected over time. However, the time origin, such as time of disease/impairment onset, is often uncertain. In order to model non-linear decline patterns in functional test scores and associated risk factors with such data, we propose two approaches as alternatives to Liu et al. One approach models change over adjacent visits with varying time intervals. The second models the change since baseline using a random effect for heterogeneity of change. We used these two approaches to examine the decline in cognitive test scores among special care unit (SCU) and non-SCU residents at the New York sites of the National Institute on Aging (NIA) collaborative studies of special dementia care. The analyses suggest that, controlling for several covariates, SCU residents experienced more rapid cognitive decline than did non-SCU residents. The relative advantages and disadvantages of the two models are discussed.

Psychological versus pharmacological treatments of bulimia nervosa: predictors and processes of change.

This article extends the acute outcome findings from a study comparing psychological and pharmacological interventions for bulimia nervosa (B.T. Walsh et al., 1997) by examining 3 additional domains: predictive factors, therapeutic alliance, and time course of change. One hundred twenty women were randomized to cognitive-behavioral therapy (CBT), supportive psychotherapy (SPT) plus antidepressant medication or a placebo, or a medication-alone condition. Results indicate that high baseline frequencies of binge eating and vomiting, as well as a positive history of substance abuse or dependence, are negative prognostic indicators. Although a greater overall therapeutic alliance may increase the likelihood of remission, symptom change over the course of treatment may have as much of an impact on patient ratings of alliance as the reverse. CBT was significantly more rapid than SPT in reducing binge eating and vomiting frequencies.

Toward a clinical model of suicidal behavior in psychiatric patients.

OBJECTIVE: Risk factors for suicide attempts have rarely been studied comprehensively in more than one psychiatric disorder, preventing estimation of the relative importance and the generalizability of different putative risk factors across psychiatric diagnoses. The authors conducted a study of suicide attempts in patients with mood disorders, psychoses, and other diagnoses. Their goal was to determine the generalizability and relative importance of risk factors for suicidal acts across diagnostic boundaries and to develop a hypothetical, explanatory, and predictive model of suicidal behavior that can subsequently be tested in a prospective study. METHOD: Following admission to a university psychiatric hospital, 347 consecutive patients who were 14-72 years old (51% were male and 68% were Caucasian) were recruited for study. Structured clinical interviews generated axis I and axis II diagnoses. Lifetime suicidal acts, traits of aggression and impulsivity, objective and subjective severity of acute psychopathology, developmental and family history, and past substance abuse or alcoholism were assessed. RESULTS: Objective severity of current depression or psychosis did not distinguish the 184 patients who had attempted suicide from those who had never attempted suicide. However, higher scores on subjective depression, higher scores on suicidal ideation, and fewer reasons for living were reported by suicide attempters. Rates of lifetime aggression and impulsivity were also greater in attempters. Comorbid borderline personality disorder, smoking, past substance use disorder or alcoholism, family history of suicidal acts, head injury, and childhood abuse history were more frequent in suicide attempters. CONCLUSIONS: The authors propose a stress-diathesis model in which the risk for suicidal acts is determined not merely by a psychiatric illness (the stressor) but also by a diathesis. This diathesis may be reflected in tendencies to experience more suicidal ideation and to be more impulsive and, therefore, more likely to act on suicidal feelings. Prospective studies are proposed to test this model.

Medication and psychotherapy in the treatment of bulimia nervosa.

OBJECTIVE: Two treatments for bulimia nervosa have emerged as having established efficacy: cognitive-behavioral therapy and antidepressant medication. This study sought to address 1) how the efficacy of a psychodynamically oriented supportive psychotherapy compared to that of cognitive-behavioral therapy; 2) whether a two-stage medication intervention, in which a second antidepressant (fluoxetine) was employed if the first (desipramine) was either ineffective or poorly tolerated, added to the benefit of psychological treatment; and 3) if the combination of medication and psychological treatment was superior to a course of medication alone. METHOD: A total of 120 women with bulimia nervosa participated in a randomized, placebo-controlled trial. RESULTS: Cognitive-behavioral therapy was superior to supportive psychotherapy in reducing behavioral symptoms of bulimia nervosa (binge eating and vomiting). Patients receiving medication in combination with psychological treatment experienced greater improvement in binge eating and depression than did patients receiving placebo and psychological treatment. In addition, cognitive-behavioral therapy plus medication was superior to medication alone, but supportive psychotherapy plus medication was not. CONCLUSIONS: At present, cognitive-behavioral therapy is the psychological treatment of choice for bulimia nervosa. A two-stage medication intervention using fluoxetine adds modestly to the benefit of psychological treatment.

Thought disorder in adolescent-onset schizophrenia.

The nature of the thinking disturbances found in adolescent-onset psychotic conditions is not as well-characterized as the thought disorders found in adult psychotic patients. We used the Thought Disorder Index to examine whether schizophrenic patients in whom psychotic symptoms appear in adolescence show the same characteristic features of thought disorder as do adult schizophrenics. Quantitative and qualitative features of thought disorder were assessed in psychiatric inpatients with adolescent-onset schizophrenia, psychotic depression, and nonpsychotic conditions compared with normal control adolescents. Elevated thought disorder occurred in all groups of adolescents hospitalized for an acute episode of psychiatric illness. The magnitude of the elevation and the frequency of occurrence of disordered thinking were greatest in the psychotic adolescents. The qualitative features of the thought disturbances found in the schizophrenic adolescents were distinct from those observed in adolescents with psychotic depression. The thinking of the schizophrenic adolescents resembled that of adult schizophrenics. In both conditions thought disorder is characterized by idiosyncratic word usage, illogical reasoning, perceptual confusion, loss of realistic attunement to the task, and loosely related ideas.

Quantifying the speed of symptomatic improvement with electroconvulsive therapy: comparison of alternative statistical methods.

Although electroconvulsive therapy (ECT) is believed to have a rapid onset of antidepressant activity, there has been limited investigation in this area. This study contrasted alternative statistical methods for testing treatment group differences in the rapidity of clinical response to ECT. Patients with major depression were randomly assigned to receive right unilateral or bilateral ECT and low or high electrical dosage relative to seizure threshold. The 24-item Hamilton Rating Scale for Depression (HRSD) was administered by blinded clinical raters twice weekly (non-treatment days). We evaluated four alternative statistical strategies. Two methods considered time to improvement as a dependent variable: (a) time (treatment number) to reach various cutoffs for percentage decrease in HRSD from baseline; and (b) survival analysis using the same cutoffs for percentage decreases as endpoints. Two methods considered time to improvement as an independent variable: (c) the slope of linear regression of HRSD scores against treatment number; and (d) a random regression model using the HRSD scores as repeated measures. The statistical methods differed in whether or not omnibus group differences were observed, the criterion level of improvement associated with group differences, and the results of pairwise comparisons establishing specific therapeutic advantages. Survival analysis generally displayed the greatest sensitivity in detecting treatment group differences.

Proton magnetic resonance spectroscopy of the temporal lobes in schizophrenics and normal controls.

Previous research has found structural and functional abnormalities in the temporal lobes of schizophrenic patients, often with greater impairment on the left side. This study applied proton MRS to both right and left temporal lobes of schizophrenic patients and normal control subjects. Reductions in the NAA/Cr ratio were found bilaterally for schizophrenic patients as compared to normal controls, and may be associated with reduced neuronal integrity. These results strengthen the evidence for biochemical abnormalities in the temporal lobes in schizophrenia.

Functional magnetic resonance imaging of schizophrenic patients and comparison subjects during word production.

OBJECTIVE: This study was undertaken to test the feasibility of using functional magnetic resonance imaging (MRI) to examine changes in cortical activation in response to verbal tasks in two brain regions. METHOD: Twelve schizophrenic patients and 11 comparison subjects underwent functional MRI of the frontal and temporal lobes. Stimulus sequences were divided into five 30-second segments by using a task-activation paradigm that alternated between resting and stimulated states. Functional images were collected every 30 seconds by using a gradient echo pulse sequence. RESULTS: Schizophrenic subjects demonstrated significantly less left frontal activation and greater left temporal activation than comparison subjects during a word fluency task. CONCLUSIONS: These preliminary data suggest that functional MRI has the sensitivity to detect differences in activation between comparison subjects and schizophrenic patients during higher cortical functions. These findings are in agreement with PET studies that reported reduced left frontal activation during challenge paradigms for the schizophrenic patients.

A double-blind dose-reduction trial of fluphenazine decanoate for chronic, unstable schizophrenic patients.

OBJECTIVE: This study evaluated the feasibility and impact of gradually reducing relatively high doses of fluphenazine decanoate by one-half for chronically impaired, poor inner-city patients with schizophrenia. METHOD: Forty-three patients currently receiving an average of 23 mg (0.3 mg/kg) of fluphenazine decanoate every 2 weeks were divided alternately into a group to remain at current doses (control group) and a group to undergo stepwise 50% dose reduction over 5 months under double-blind conditions. Clinical status and side effects were assessed quarterly for a year. Relapse was determined clinically and by changes in psychopathology ratings. RESULTS: Eighty-six percent (N = 37) of the patients (control group, N = 17; reduced-dose group, N = 20) completed the study. The groups did not differ at baseline in demographic or clinical variables or neuroleptic dose. In the reduced-dose group, doses were lowered to an average of 11.5 mg every 2 weeks. The two groups did not differ throughout the year in number of relapses, and hospitalization rates fell similarly in both (overall, by about 67%). Clinical measures changed little. Extrapyramidal symptoms worsened in the control group but improved slightly in the reduced-dose group. Tardive dyskinesia worsened in both groups, but less in the reduced-dose group. CONCLUSIONS: Maintenance neuroleptic doses much lower than the conventional ones can be achieved safely in schizophrenic patients by gradual reduction, without clinical worsening and perhaps with fewer extrapyramidal symptoms and less tardive dyskinesia. The two-thirds lower hospitalization rate, with substantial financial savings, apparently was due to nonspecific effects of research intervention.

One hundred years of schizophrenia: a meta-analysis of the outcome literature.

OBJECTIVE: This study was undertaken to assess the twentieth-century literature on outcome in schizophrenia for historical trends that might be associated with changes in diagnostic and therapeutic practice and to test the hypothesis that both improved biological treatment and changes in diagnostic criteria have influenced outcome. METHOD: Meta-analysis of the international literature on outcome in schizophrenia or dementia praecox from 1895 to 1992 identified 821 studies; 320 of these, with 51,800 subjects in 368 cohorts, met the inclusion criteria for the study. RESULTS: Only 40.2% of patients were considered improved after follow-ups averaging 5.6 years (range = 1-40). Outcome was significantly better when patients were diagnosed according to systems with broad criteria (46.5% were improved) or undefined criteria (41.0% were improved) rather than narrow criteria (27.3% were improved). The proportion of patients who improved increased significantly after mid-century (for 1956-1985 versus 1895-1955, 48.5% versus 35.4%), probably reflecting improved treatment as well as a broadened concept of schizophrenia. However, in the past decade, the average rate of favorable outcome has declined to 36.4%, perhaps reflecting the re-emergence of narrow diagnostic concepts. CONCLUSIONS: Overall, less than half of patients diagnosed with schizophrenia have shown substantial clinical improvement after follow-up averaging nearly 6 years. Despite considerable gains in improvement rates after mid-century, there has been a decline since the 1970s. These historical changes probably reflect improved treatment, shifts in diagnostic criteria, and selection bias related to changes in health care.

Obstetrical complications and trail making deficits discriminate schizophrenics from unaffected siblings and controls.

Numerous studies have reported that both obstetrical complications (OCs) and deficits on the Trail Making Test show elevated prevalences in schizophrenics. Trail Making deficits have also been reported to be more common in schizophrenics' relatives than in controls, suggesting poor Trail Making performance may be a behavioral indicator of a familial risk factor for schizophrenia. Few studies, however, have investigated how these two variables co-vary in samples of schizophrenics and non-schizophrenics. In this study, DSM-III-R diagnoses, OCs noted in birth records, and Trail Making performance were independently assessed in 30 subjects: 9 schizophrenics, 8 of their non-schizophrenic siblings, and 13 comparison subjects with neither a personal nor a family history of schizophrenia. Results supported two key predictions of a two-factor etiologic model of schizophrenia: (a) the combination of perinatal OCs and poor Trail Making performance discriminated schizophrenics extremely well from non-schizophrenics, including their own non-schizophrenic sibs, and (b) perinatal OCs and Trail Making errors manifested a significant inverse association among schizophrenics' non-schizophrenic sibs, but not among other subjects.

Empirical determination of thresholds for case identification: validation of the Personality Diagnostic Questionnaire-Revised.

The Personality Diagnostic Questionnaire-Revised (PDQ-R) was sent to first-degree relatives of major psychotic patients for identification of DSM-III-R personality disorders (PDs). Responses to the PDQ-R were interpreted both literally and empirically, and compared with the Structured Interview for DSM-III PDs (SIDP) as the standard. For literal interpretation, symptoms reported were counted directly for case identification using fixed DSM-III-R thresholds. The empirical approach adjusted the threshold for case identification to maximize concordance with the SIDP. Comparison of the two methods showed that using empirically determined thresholds in some scales gives better concordance with the SIDP. For the dependent and histrionic PD scales, the improvements were statistically significant. The area under the receiver operating characteristic (ROC) curve was computed for each PDQ-R scale to summarize its discriminatory capability across all thresholds. Areas under the ROC curve indicated that the schizoid, schizotypal, borderline, dependent, passive-aggressive, and histrionic PD scales in the PDQ-R have better discriminatory qualities than other PDQ-R scales.

Some conceptual and statistical issues in analysis of longitudinal psychiatric data. Application to the NIMH treatment of Depression Collaborative Research Program dataset.

Longitudinal studies have a prominent role in psychiatric research; however, statistical methods for analyzing these data are rarely commensurate with the effort involved in their acquisition. Frequently the majority of data are discarded and a simple end-point analysis is performed. In other cases, so called repeated-measures analysis of variance procedures are used with little regard to their restrictive and often unrealistic assumptions and the effect of missing data on the statistical properties of their estimates. We explored the unique features of longitudinal psychiatric data from both statistical and conceptual perspectives. We used a family of statistical models termed random regression models that provide a more realistic approach to analysis of longitudinal psychiatric data. Random regression models provide solutions to commonly observed problems of missing data, serial correlation, time-varying covariates, and irregular measurement occasions, and they accommodate systematic person-specific deviations from the average time trend. Properties of these models were compared with traditional approaches at a conceptual level. The approach was then illustrated in a new analysis of the National Institute of Mental Health Treatment of Depression Collaborative Research Program dataset, which investigated two forms of psychotherapy, pharmacotherapy with clinical management, and a placebo with clinical management control. Results indicated that both person-specific effects and serial correlation play major roles in the longitudinal psychiatric response process. Ignoring either of these effects produces misleading estimates of uncertainty that form the basis of statistical tests of hypotheses.

A family study of self-reported personality traits and DSM-III-R personality disorders.

Personality traits and DSM-III-R personality disorders of first-degree relatives of patients with psychoses were studied using the NEO Five-Factor Inventory (NEO-FFI) and the Personality Diagnostic Questionnaire-Revised (PDQ-R), two self-report instruments. The relatives were compared on their scores for the five personality factors in the NEO-FFI, the prevalence of individual DSM-III-R personality disorders, and their scores for each personality disorder scale in the PDQ-R. The results suggest that there is no difference in personality traits and prevalence of personality disorders, including schizophrenia spectrum disorders, when relatives of patients with schizophrenia, bipolar disorder, and major depression are compared to each other. Relatives of patients with "atypical psychosis," psychotic disorders that do not meet DSM-III-R criteria for any specific nonorganic psychotic disorder, may be a distinctive group.

The relationship between DSM-III personality disorders and the five-factor model of personality.

Two hundred twenty-four first-degree relatives of patients with psychotic disorders were administered the Structured Interview for DSM-III Personality Disorders (SIDP) and completed a self-report instrument to assess dimensions of the five-factor model of personality. All of the DSM-III personality disorders were related to one or more dimensions of the five personality factors; however, the correlations were generally low. It seems that the five personality factors describe important features of DSM-III personality disorders, but are not sufficient to completely explain their characteristics. Future use of the five-factor model in conjunction with personality disorder diagnoses may provide useful information for clinical work and research purposes.

Haloperidol response and plasma catecholamines and their metabolites.

Eleven acutely psychotic patients with schizophrenia or schizoaffective disorder underwent a 5-7 day drug-washout period (with lorazepam allowed) prior to participating in a 6-week controlled dose haloperidol trial. Patients were evaluated longitudinally with clinical ratings and with plasma measures of the catecholamines dopamine (pDA) and norepinephrine (pNE) and their metabolites, homovanillic acid (pHVA) and 3-methoxy-4-hydroxyphenylglycol (pMHPG). All patients exhibited clinical improvement with haloperidol; the decrease in their Brief Psychiatric Rating Scale (BPRS) scores ranged from 32 to 89%. Measures of pHVA increased within the first week of treatment and returned to baseline by week 5. The pattern of change of pDA resembled that of pHVA. The pattern of change of pNE and pMHPG revealed a decrease over the course of treatment. The early increase and the subsequent decrease in pHVA were strongly correlated with improvement in positive symptoms on the BPRS. These data are consistent with previous reports on the change in pHVA and pMHPG during clinical response to haloperidol. The data on change of pDA and pNE further describe the nature of the biochemical response to this drug.

Clozapine response and plasma catecholamines and their metabolites.

The atypical neuroleptic clozapine has an unusual profile of clinical effects and a distinctive spectrum of pharmacological actions. Plasma measures of catecholamines and their metabolites have been used in the past to study the action of typical neuroleptics. We obtained longitudinal assessments of plasma measures of dopamine (pDA), norepinephrine (pNE), and their metabolites, homovanillic acid (pHVA) and 3-methoxy-4-hydroxyphenylglycol (pMHPG), in eight treatment-resistant or treatment-intolerant schizophrenic patients who were treated with clozapine for 12 weeks following a prolonged drug-washout period. Our findings from the study of these eight patients suggest the following: Plasma levels of HVA and possibly NE derived from the neuroleptic-free baseline period may predict response to clozapine; plasma levels of HVA and MHPG decrease during the initial weeks of treatment in responders but not in nonresponders; and plasma levels of DA and NE increase in both responders and nonresponders to clozapine.