International Biometric Society
British and Irish Region

Statistical Aspects of Safety Data

London
Wednesday June 28^th 2006

Current issues in drug safety are addressed (such as QT prolongation and the use of Bayesian methods), as well as issues arising from food safety work. The meeting will be of interest to statisticians and other researchers who work in these areas and aims to promote a wider cross-fertilisation of ideas as applied to safety signal detection.

Advance registration is required for this meeting, with the costs of registration doubling after May 24^th. The registration cost (£52.88 (incl. VAT) for IBS/PSI members, £88.13 (incl. VAT) for non-members before 24^th May; £105.75 (incl. VAT) for IBS/PSI members, £176.25 (incl. VAT) for non-members after 24^th May) includes lunch and coffee/tea during the meeting.

A one-page flier for the meeting is attached, together with a registration form. Completed registration forms can be faxed to Alison Houghton on 01625 267879, or sent by post to:

Please note that it is also possible to register and pay for this meeting online at www.psiweb.org.

If you have any queries about the meeting please contact Peter Colman, IBS British and Irish Region Committee member responsible for this joint meeting, at peter.colman@pfizer.com.

Programme and Abstracts

09.15 – 09.45	Registration
09.45 – 10.25	Mechanistic basis of adverse drug reactions: the perils of inappropriate dose schedules
	Rashmi Shah, Independent Consultant (ex-MHRA)
	Adverse drug reactions (ADRs) have long been recognised as a significant cause of morbidity and mortality. By far the most common ADRs are the concentration dependent pharmacological reactions, the majority of which ought to be preventable. Dose schedules investigated during clinical development programmes suffer from a number of deficiencies. Withdrawal of a number of drugs, particularly cerivastatin and cisapride, illustrates the point. Historical review of dose schedules of a wide variety of drugs confirms that the trend to inappropriate dose selection has changed little over the last 35 years. Apart from high doses, a number of other factors lead to high plasma concentrations of parent drug or its metabolites. Better drug development strategies and increased awareness and education of prescribers and pharmacists offer great opportunities for substantially minimising concentration-related ADRs.
10.25 – 11.05	Probabilistic exposure assessment in food safety
	Andy Hart, Central Science Laboratory (York)
	Monte Carlo simulations are increasingly popular for modelling exposure of consumers to nutrients, contaminants and pathogens in food. Most attention has been given to quantifying variation in exposure, by combining distributions for consumption and concentration. It is also important to characterise the uncertainties affecting exposure estimates, but these are more difficult to quantify. Examples involving pesticides and chemicals from food packaging will be used to illustrate current approaches and outstanding challenges.
11.05 – 11.25	Coffee
11.25 – 12.05	Detecting the presence of contaminants and unauthorised ingredients in foods.
	Roy MacArthur, Central Science Laboratory (York)
	The detection of contaminants and unauthorised ingredients in foods is necessary to ensure the safety and authenticity of food products. A collection of methods is presented for estimating the effect of heterogeneity of the sampling target on the ability to detect and quantify contaminants. The methods focus on answering the questions “How much contaminant might be present in the sampling target given a negative test result” and “How can high quality measurement results be produced most efficiently?”
12.05 – 12.35	Drug-induced QT interval prolongation: its significance and limitations during drug development – a clinical and a statistical perspective
	Rashmi Shah, Independent Consultant (ex-MHRA) and Georg Ferber, Novartis
	RS: A QTc interval is a measure of time between two identifiable points of your heart-beat on an ECG. Prolongation of the QTc interval, either congenital or drug-induced, can result in potentially fatal ventricular arrhythmia known as torsade de pointes. This property has been responsible in recent times for the withdrawal of many drugs from the market and the number of such drugs continues to increase almost daily. QT interval prolongation has been regarded as a liability and regulatory authorities have rejected many new drugs or placed restrictions on the use of many old and new drugs. However, QTc interval prolongation is an imperfect surrogate of the clinical risk and a number of factors modulate the clinical risk of torsade. Therefore, although evaluating the effect of an NCE on QTc interval is important, conclusions on the potential clinical risk of torsade associated with its use, based solely on its ability to prolong the QTc interval, might turn out to be highly flawed.
	GF: The biggest statistical challenge of the ICH E14 guideline on the assessment of proarrhythmic potential of non-antiarrhythmic drugs is the design of the "Thorough QT/QTc Study", which is one of the rare cases of a confirmatory safety trial. The statistical discussion during the development of this guidance centred on the definition of the "largest time matched mean effect", and on a pragmatic way to ensure that this effect does not exceed a threshold value of around 5 ms. Identifying an intersection-union test procedure based on non-inferiority tests at each of a series of time points and understanding its basic properties was one of the prerequisites for promoting the guideline to step 4 and then make it effective last year. I will give an overview over the key concepts involved.
12.35 – 13.35	Lunch
13.35 – 14.15	Relating data from the in vitro hERG assay and from dog models to QT interval prolongation in humans
	Daniel Jonker, Grünenthal GmbH
	Early on in drug development, an adequate prediction of drug-induced QT interval prolongation in humans is essential for the selection of candidate drugs. A population PK/PD analysis is presented that integrates data on the antiarrhythmic drug dofetilide obtained from an in vitro hERG assay, a dog model and QT studies in humans. Insight is provided into the relationships of the responses in the preclinical assays to the concentration-QT curve in humans and into the precision for predicting small QT effects.
14.15 – 14.45	Optimising the design of thorough QT/QTc studies
	Monica Antunes, Novartis
	Thorough QT/QTc studies are conducted with the aim of assessing the effect of a study drug on the prolongation of the QT/QTc intervals. Planning such a study requires fundamental decisions to be made concerning the choice of the design. This talk will investigate the optimisation of the study design under statistical, ethical and economical considerations based on some case studies. In particular, the choice of baseline measurements and the value of replicates will be discussed.
14.45 – 15.05	QT interval prolongation – discussion
15.05 – 15.25	Coffee
15.25 – 16.05	Are Bayesian methods the solution to safety signals in trials and pharmacovigilance?
	Stephen Evans, London Schoolof Hygiene and Tropical Medicine
	A fundamental problem in dealing with issues of finding harms associated with medicines is that of multiplicity. At the same time, in contrast to evaluating efficacy, the clinical problem of most relevance is failing to find real effects. Standard frequentist approaches to multiplicity tend to have high type II errors and the recent developments in analysing spontaneous reporting databases (duMouchel and Bate) have proved useful. This talk will review the methods and their evaluation and make some suggestions for further development and research.
16.05 – 16.45	Bayesian methods in safety assessment
	Deborah Ashby, Wolfson Institute of Preventive Medicine
	Safety data have many features that make them difficult to study, and a Bayesian approach may have something to offer. After outlining these arguments, a recent systematic review carried out with two colleagues will be presented. The review focuses on pharmaco-epidemiological studies and maps the literature on 'Bayesian safety' covering methodology, illustrative applications, and examples where a Bayesian approach was the main technology in real applications. Case-studies from the review will be presented in detail. The talk will finish by pointing to fruitful areas for application of Bayesian methods and the potential for further research in the area.