Please activate JavaScript!
Please install Adobe Flash Player, click here for download

ISCB2014_abstract_book

ISCB 2014 Vienna, Austria • Abstracts - Oral Presentations 65Wednesday, 27th August 2014 • 9:00-10:48 Monday25thAugustTuesday26thAugustThursday28thAugustAuthorIndexPostersWednesday27thAugustSunday24thAugust C35 The biostatistician’s toolbox II C35.1 Pharmacodependence: new graphical representations E Peron1,2 , F Feuillet2,3 , R Morgane1 , C Victorri-Vigneau1,2 , J-B Hardouin2,3 1 Service de Pharmacologie Clinique, Nantes, France, 2 EA 4275 SPHERE, Nantes, France, 3 Plateforme de Biométrie, Nantes, France   Introduction: In France, a network of 13 Centers for Evaluation and Information on Pharmacodependence (CEIP) monitors substance abuse risk. This network is coordinated by the French medicines agency. To assess abuse and dependence potential of drugs, CEIPs record cases of substance abuse and dependence arising from health professionals. Each substance reported is evaluated by the pharmacodependence grav- ity score. This is an 8-item score (each item is binary rated: positive or negative) di- vided into 2 different paths: • The first path assesses physical and compulsive signs: tolerance, withdrawal syndrome, dose taken in larger amounts or over a longer period than was originally expected, desire to cut down. • The second path assesses harmful consequences of the pharmacodependence: a great deal of time is spent, interpersonal problems, consummation persistence despite health problems, transgression behavior. At the end of this assessment, we obtain for each substance a score on 8 consists of 2 sub-scores on 4 for each path. Methods: This work aimed to develop new graphical representations and indexes to compare substances’ pharmacodependence profile using the CEIP electronic records. For instance, 3 substances which have very different pharmacodepen- dence profile will be shown: buprenorphine, heroine and a control: parox- etine - a non-potential pharmacodependence substance. Results: Graphical representations and indexes allow distinguishing be- tween substances using the score and the 2 sub-scores. Their interpreta- tion will be explained. Conclusion: The distinguished feature of the pharmacodependence score and the 2 sub-scores is demonstrated.This work open up new prospects of methodological tools development. C35.2 Constructing robust confidence intervals for drug utilization time series data A Corkum1 , Y Zhang1 , P Cabilio1 1 Acadia University, Wolfville, Canada   For a drug safety and effectiveness study, we are interested in constructing confidence intervals in drug utilization time series to investigate the level/ trend over a signal or several time periods. The current methods either as- sume that the data follow an independent process, or rely on the explicit knowledge of underlying process distribution and its dependence struc- ture, which is not applicable. In this paper, we generalize properties of the sign and Wilcoxon signed rank statistics to time series data, and develop new nonparametric meth- ods to construct confidence intervals when the data are dependent. Specifically, we use the theory of U-statistics for mixing processes to de- velop asymptotic normality theorems for these statistics, which are then used to approximate the confidence interval.The variances of the sign and Wilcoxon signed rank statistics are keys in computing such confidence in- tervals. We consider both the unrealistic case in which we know the underlying process distribution and its dependence structure and are able to com- pute the variances exactly, and the more realistic situation in which the variance must be estimated due to lack of information. We implement three methods for the variance estimation: block bootstrapping; sieve bootstrapping; and the empirical distribution method, and then com- pare them with the exact confidence interval coverage through extensive simulations. The results show that the proposed methods are effective and robust for time series data. Finally, we illustrate our methods with time series data on a heartburn medication to investigate the effect of a provincial insurance policy change.   C35.3 Using constraints to compare state structures in cost-effectiveness decision models H Thom1 , C Jackson2 , N Welton1 , L Sharples3 1 University of Bristol, Bristol, United Kingdom, 2 MRC Biostatistics Unit, Cambridge, United Kingdom, 3 University of Leeds, Leeds, United Kingdom   Cost-effectiveness decision models are used to estimate expected costs and effects of interventions for the management of disease and thus guide the decision making of national health services. These are often multistate models with states corresponding to categorizations of disease status. A common difficulty is the choice of states for the multistate model and it may be unclear whether there are sufficient data to inform the tran- sition probabilities. Similar states could be merged but different structures can give different decision recommendations. Our aim is to compare different structures by balancing fit and parsimony based on the available data, and to quantify the associated decision un- certainty in terms of the expected value of perfect information about the model and its parameters. Models with different states are difficult to com- pare by standard statistical methods because they are fitted to different data and the corresponding likelihoods are on different scales. However, we will show that models with coarsened state structures are practically equivalent to special cases of a single, sufficiently flexible, mod- el.These special cases are defined by constraints on the model parameters and can be compared statistically as they are fitted to the same data. The expected value of perfect information of the extra parameters required to relax the constraints therefore represents the decision uncertainty associ- ated with the model structure. We shall illustrate this approach for a variety of state transition patterns used in cost-effectiveness models, and present an application to a model for diagnostic testing strategies for coronary artery disease. C35.4 Optimal and maximin sample sizes for multicentre cost-effectiveness trials A Manju1 , MJJM Candel1 , MPF Berger1 1 Maastricht University, Maastricht, The Netherlands   This paper deals with the optimal sample sizes for a multicentre trial in which the cost-effectiveness of two treatments in terms of net-monetary benefit is studied. The optimal sample sizes concern the number of cen- tres and the number of individuals per centre in each of the treatment conditions. These numbers maximize the efficiency or power for given research costs or minimize the research costs at a desired level of efficiency or power. Information on several model parameters of a bivariate linear mixed mod- el and sampling costs are required to calculate these optimal sample sizes. In case of limited information on relevant model parameters, sample size formulas are derived for so-called maximin sample sizes which guarantee a desired power level at the lowest study costs. Four different maximin sample sizes were derived based on signs of the lower bounds of the cor- relation between random slopes for costs and effects and individual level correlation between costs and effects, where one case is worst compared

Pages Overview