Abstracts from ASENT 2004 Annual Meeting March 11–13, 2004

Multiple Imputation Compared with Conventional Methods for Analysis of Categorical Repeated Measures 1 6 10 Yuyan Duan 1 2 0 6 7 10 11 14 17 Ilya Lipkovich 0 1 3 4 6 8 10 12 16 Saeeduddin Ahmed 0 1 3 4 6 8 10 12 16 0 Lilly Research Laboratories , Indianapolis, Indiana 1 Saeeduddin Ahmed, Ilya Lipkovich, Jonna Ahl, Thomas Hardy, Diane Haldane, Robert Baker, and Mauricio Tohen 2 Virginia Polytechnic Institute and State University , Blacksburg, Virginia 3 Department of Psychiatry, University of Pittsburgh , Pittsburgh, Pennsylvania 4 Psychiatry and Bioethics Program, University of Michigan , Ann Arbor, Michigan 5 Department of Neurology and Otorhinolaryngology, University of Rome La Sapienza , Rome, Italy 6 Department of Neurology, The University of Chicago , Chicago, Illinois 7 Department of Neurology, University of Rochester , Rochester, New York 8 Brigham and Women's Hospital, Harvard Medical School , Boston, Massachusetts 9 Huntington Medical Research Institute , Pasadena, California 10 James P. Bennett, Jr. Center for the Study of Neurodegenerative Diseases, Department of Neurology, University of Virginia , Charlottesville, Virginia 11 Pepgen Corporation, Alameda, California 12 Center for Neural Recovery and Rehabilitation Research, Helen Hayes Hospital and Columbia University , West Haverstraw, New York 13 Department of Neurosurgery, Columbia University School of Medicine , New York, New York 14 Maryland Psychiatric Research Center, University of Maryland School of Medicine , Baltimore, Maryland 15 Department of Neurology, University of Virginia , Charlottesville, Virginia 16 Department of Psychology, Central Michigan University , Mount Pleasant, Michigan 17 Experimental Therapeutics Branch, National Institutes of Health , Bethesda, Maryland Background: Analyses of categorical repeated measures of clinical data using conventional methods can give biased estimates of treatment effects and associated SEs when dropouts are not completely at random (depending on observed clinical outcomes). We test the utility of multiple imputation (MI) analysis in reducing these biases. Methods: We used simulation to compare performance of MI versus conventional methods, including restricted pseudolikelihood methods and generalized estimating equations, in five typical clinical profiles for 1) estimating overall treatment effects, and 2) treatment differences at last scheduled visit. Results: The power to detect treatment differences with MI is consistently higher than with conventional methods. Type I error rates (estimated from scenarios in which no treatment difference existed) were consistently smaller with MI than with conventional methods. However, MI tended to overestimate variability of treatment differences at endpoint. Among tested profiles, the advantage of MI over conventional methods in terms of power to detect overall treatment differences was greatest when treatments separated from each other early, then converged later. Conclusion: Compared to conventional techniques, MI may lead to less biased estimates of treatment differences in categorical analyses of continuous data, especially in clinical trials with a high (40-60%) proportion of dropouts. However, MI did not perform well when dropouts were (partially) driven by clinical outcomes that were also not observed. Of course, this conclusion is limited by the specifics of the simulation scenarios tested and as such, does not constitute theoretical proof. This work was supported by Eli Lilly and Company. Background: Atypical antipsychotics are being used as mood stabilizers for treatment of bipolar disorder. Weight gain, sometimes substantial, is observed during treatment with some of the atypical agents. Objective: To examine early weight gain (EWG) during olanzapine (OLZ) treatment as a predictor of substantial weight gain (SWG) later in treatment. Methods: Data were pooled from seven randomized, multicenter studies in bipolar patients (n 2231 at initiation of olanzapine). Endpoint for this analysis was 32 weeks (n 518). SWG was defined as gaining at least 10 kg at endpoint. Other definitions of SWG (5 and 7 kg) were also evaluated. EWG was determined as the weight gain threshold that optimized discrimination of patients with SWG. Results: At an EWG threshold of about 2 kg in the first 3 weeks, among patients with SWG 61% also had EWG (sensitivity), and 69% of patients without SWG did not have EWG (specificity). Among those patients who did not have EWG, 83% did not have SWG. Addition of baseline characteristics (including age, gender, baseline body mass index, nonwhite race) moderately improved predictability. Results for other SWG cut-offs were qualitatively similar. Conclusions: EWG of about 2 kg during the first 3 weeks of treatment with OLZ predicted later SWG. Patients with lesser EWG appeared much less likely to have SWG. Further research in larger patient populations and for longer time periods is necessary to explore the predictive power of EWG for SWG with olanzapine as well as other medications, particularly in patients with symptomatic bipolar disorder. - Early Weight Gain as a Predictor of Substantial Weight: Results from Seven Studies of Olanzapine for the Treatment of Bipolar Disorder Introduction: This research was conducted to better understand the phenomenon of antipsychotic treatment discontinuation and to provide insight into overall treatment effectiveness. Methods: This was a post hoc, pooled analysis based on four randomized, double-blind clinical trials that had a duration of 24-28 weeks. The four studies included 822 olanzapine-treated patients and 805 patients treated with risperidone, quetiapine, or ziprasidone. Results: Adverse events related to worsening in psychiatric symptoms accounted for 50% (100 of 200) of all adverse events leading to treatment dropouts, with worsening in psychotic disorder (35 of 200) and suicide (completed/attempt/ideation, 18 of 200) being the most frequent events. Most decisions of discontinuation due to lack of efficacy were made based on patient perception (132 of 164). Clinical rating scales showed patients who discontinued because of lack of efficacy received no relief of their psychiatric symptoms. Conclusions: This research demonstrated that worsening of underlying psychiatric symptoms as well as patients perception of their failure to improve overwhelmingly contribute to treatment discontinuation, which can threaten patient wellbeing with the morbid consequences of illness exacerbation. Better understanding of what causes discontinuation may provide a strategy to improve patient engagement in long-term therapy and to increase patient access in realizing the goals of an effective treatment. This work was supported by Eli Lilly and Company. Comparison of Olanzapine and Lithium Treatment Groups Based on Lithium Blood Levels: A 52-Week Bipolar Relapse Prevention Study Saeed Ahmed, Yuyan Duan, Kristine Healey, Ilya A. Lipkovich, and Mauricio F. Tohen (...truncated)