The research program we propose concerns the development of clinical trial methodology, including new experimental designs and methods of statistical analysis, for the study of antipsychotic drugs used in the treatment of schizophrenia: (1) Although """"""""flexible dose"""""""" or """"""""doctor's choice"""""""" designs mirror clinical practice and accommodate individual patient variations they have been seriously criticized. We will develop statistical bioassay models that estimate relative potency while taking into account placebo responders and refractory.patients as well as the dosage information. (2) When the relative potency of a test to a standard is not the same for all behavioral and side effect variables we will develop methods that permit the identification of subgroups of the variables with respect to which relative potency is constant. (3) Generally, the categories within items in a rating scale have implicit rank order quantified by arbitrarily assigning equispaced integer values to the categories. We will develop methods whereby scalar assignments are made on the basis of dose response data from the trial itself. (4) Crossover designs that are efficient and are analyzable whether or not carryover effects are present will be developed so that a relatively small sample size will have adequate power to contrast therapies. (5) Models to analyze multivariate observations for evaluation of the bioequivalence of two treatments through the simultaneous analysis of serum levels of drug and/or metabolites at all of the observation time points will be developed. (6) We shall develop statistical methodology for assessing the merits of combination therapies. The nature of the null hypothesis has led to considerable confusion in the analysis of studies of combination therapy. These usually involve hypotheses of the form HO: combinations not better than A or combination: not better than B. (7) Other statistical problems that arise in consultation and collaborating with clinical researchers will be considered.
Showing the most recent 10 out of 23 publications
