First Channel Network Conference
8 - 11 May, 2007
Rolduc, Kerkrade, The Netherlands
|
|
First Channel Network Conference8 - 11 May, 2007
|
|
Online Registration NOW OPEN!!
Registration costs increase after March 25th
The conference will start at lunchtime on Tuesday May 8th, and finish at lunchtime on Friday May 11th, hopefully allowing sufficient travel time on both days for delegates from the British and Irish Region to travel to and from Rolduc without the need for an additional overnight stay. The conference venue is the imposing abbey of Rolduc, which was founded in 1104. The pattern for the conference will broadly follow that from the 2005 conference, with opening and closing Keynote presentations, three further invited sessions, a number of contributed oral sessions, and a poster session.
The Channel Network Steering Group, comprising representatives from all four regions (Joe Perry and Tim Cole representing the British and Irish Region) and chaired by Andrew Mead, Channel Network Coordinator, have put together an exciting and international invited scientific programme for the conference, focussed on three main topics - `High-dimensional data'; `Methods in epidemiology', and `Adaptive designs in clinical trials'. Details of the invited speakers, the titles of their presentations, and abstracts where available are given below:Groups, graphs, networks: New
strategies to handle genomic data
Ulrich Mansmann (IBE, University of Munich, Germany)
This talk will discuss recent developments in the analysis of high-dimensional genomic data.
The concept of a gene group is quite vague but often used in genomic research. It emphasises the conceptual and biological importance of the expression profile of a group of genes versus the multiple analysis of individual gene expression. Several authors contributed methodology to analyse differential expression in gene groups. Basic concepts will be presented like gene set enrichment, global tests and ontologic analysis. Looking for structure in gene groups, graphs allow an understanding of interactions between genes within a group. The use of graphs for the statistical evaluations of such structures will be discussed. Besides the construction of graphs it is of interest to compare graph structure between different biological entities. Genomic networks present more refined concepts which belong up to now to the field of Bioinformatics. Conceptually they are closely related to graphs, but have only a loose contact to statistical reasoning. The talk tries to bridge the gap between the more statistical models related to graphs and groups and the bioinformatic approaches related to networks.
Megavariate Analysis-of-Variance
Age Smilde, Johan Westerhuis, Huub Hoefsloot & Daniel Vis
(Swammerdam Institute of Life Sciences, Universiteit van Amsterdam,
The Netherlands)
In functional genomics studies frequently data are measured with an
underlying experimental design. Analyzing a univariate response then comes
down to applying ANOVA-type methods. An extension to multivariate
responses is also available, i.e. MANOVA. The latter method, however,
breaks down when going from multivariate to megavariate data:
measuring hundreds of responses. This calls for methods capable of
analyzing very many responses with only a limited number of samples.
Recently, such a method has been developed - ASCA - which is a
combination of ANOVA and Simultaneous Component Analysis (SCA). This
method allows for visualizing the important trends in megavariate
designed data. A permutation procedure gives an indication of the
significance of those trends. There are still a lot of open
statistical questions in this approach.
The ASCA methodology will be explained using examples from
metabolomics. Special attention will be paid to time-resolved
metabolomics measurements. Problems and pitfalls will be discussed.
Beyond lasso
Cajo J F ter Braak (Biometris, Wageningen UR, The
Netherlands)
Statistical
methods for the epidemiology of HIV/AIDS
Daniel Commenges
(INSERM, France)
We consider the main statistical methods that have been or could be applied to the estimation of the incidence of HIV infection. Such estimates can be useful not only to draw attention of the public authorities on the importance of the problem but also for assessing the impact of preventive policies. Direct studies of prevalence or incidence of HIV infection raise considerable problems. Backcalulation is an indirect method which allows to retrieve the incidence of HIV infection from AIDS incidence together with the knowledge of the distribution of the incubation period. This presupposes the existence of a reliable AIDS register; moreover even in countries where such registers exist the simple backcalculation method has become inapplicable because of the non-stationarity of the distribution of the incubation period mainly due to the spread of efficient treatments. More elaborate methods based on multistate models have been proposed. The availability of markers of recent infection may bring information especially in the most recent trend of the epidemics. It is also possible that dynamical epidemic models (explicative models based on differential equations) bring some contribution. Also it is to be noted that the methods should be adapted to the different contexts prevailing in Europe or the US, in Africa and in Asia.
Causal
inference in observational studies
Miguel Hernan (Harvard School of Public Health, USA)
Causal inference from complex longitudinal data with time-varying exposures requires analytic methods that appropriately adjust for time-varying confounding and selection bias. In these settings, conventional statistical methods may lead to estimates that cannot be endowed with a causal interpretation, whether the data arose in a randomized experiment or an observational study. This talk summarizes the main issues and presents examples from current topics in epidemiologic research.
Issues in combining Phase II and Phase III clinical
trials
Chris Jennison (University of Bath, UK)
There is much current interest in the concept of a "seamless transition" between Phase II and III trials, eliminating the hiatus that can occur as the Phase III study is designed and approved. It is also possible that data from both phases may be used in the analysis of the treatment or dose level selected for testing in Phase III.
I shall discuss the pros and cons of running two separate trials for Phases II and III, then focus on just what is involved if investigators decide to run a combined trial. A range of statistical issues arise when the Phase II and III data are analysed jointly: allowance must be made for multiple comparisons and a possible selection bias in Phase II data on the treatment chosen for Phase III; an efficient study design should have interim analyses to allow early elimination of poor treatments in Phase II and early stopping of the whole process as soon as an overall conclusion can be reached. I shall survey existing proposals, assess the impact on efficiency of elements of these proposals, and describe directions for future research.
Implementation of a Phase I Adaptive Clinical
Trial: A case study using Bayesian methods.
Fabian Tibaldi (Eli Lilly and Company, Belgium)
Phase I clinical trials are conducted in general to determine the dose-response curve of a new drug with respect to safety and tolerability. In particular, these studies aim to estimate the maximum tolerated dose (MTD). There is an increasing interest in speeding trials in order to put new drugs on the market as soon as possible. The goal is to select the best of small number of dose levels of a new compound and provide some good candidates at early stages of development. These treatments will be taken forward to Phase II or later trials. Adaptive designs have been intensively studied in last years and have been considered as a novel approach when we want to minimize the number of subjects/patients and to reduce the risk of serious adverse events. While adaptive designs are well-known for Phase II and seamless Phase II/III studies, however, in the Phase I arena are not so popular yet but are getting more and more explored. The existing methodology is quiet larger and is in constant evolution. We do not intent to discuss characteristics of these kinds of trials, neither the advantages nor limitations, but to provide an easy strategy to implement adaptive designs using free available software packages. We will present a case study of a Phase I trial and we will focus on the practical advantages of a Bayesian approach.
Fast And Sensitive Trials – New Paradigms in Clinical Research
Marc Buyse (IDDI, Louvain-la-Neuve, Belgium)
Recent developments in biostatistics and bioinformatics have reshaped the landscape of clinical research. With the number of promising new molecules available for clinical testing, trials need to detect a drug’s benefit (and harm) as fast as possible. This is possible using (group) sequential designs, which are typically monitored by Independent Data Monitoring Committees. Interim analyses are often very informative, and flexible designs purport to use the available information in a statistically controlled manner to maximize the trial’s chances of success, or conversely to terminate it as soon as possible. The need for speed and the advances in molecular biology have also led to the development of biomarkers which might be used as surrogates for clinical endpoints, if their predictive ability could be statistically established. Finally, high throughput technologies have made it possible to identify molecular profiles with exquisite prognostic and / or predictive ability. These profiles can either be used to design trials with an upfront focus on responding patients, or developed and validated during the trial itself to narrow the patient eligibility dynamically. All these developments raise formidable statistical challenges which will be discussed theoretically, and illustrated with examples of actual on-going trials.
In addition to these keynote and invited sessions we are also hoping to host a `Biometrics Showcase' session to highlight some recent publications in our journal by members of the Channel network regions.
For the rest of the conference we have attracted a lively series of contributed oral sessions, together with a poster session. Further contributions of poster presentations are still welcomed on any biometrical topic, with contributions particularly encouraged on the topics identified for the invited programme. Abstracts should be submitted by March 25th, written in English, and not exceeding one page of A4 in length (minimum font size 12pt). Each abstract should include the title, list of authors and author affiliations, and email contact for the presenting author.
Abstracts can be submitted either through the conference web-site:
or by email (either as a Word file or plain text attachment) to the chair of the local organising committee, Hein Putter, at:
Registration for the conference is now possible electronically via the conference web-site (see above), or by returning the registration form (downloadable from the conference web-site) to:
Hein PutterThe costs of registration are: This registration fee covers all meals (breakfast, lunch, dinner)
plus coffee and tea during the conference, and the conference
materials, programme, book of abstracts), but not accommodation.
Note that
the differential between the costs for IBS members and non-members is
greater than the annual subscription for Associate and Student members
(currently equivalent to about 22 euros). To qualify for the IBS member’s
registration level, delegates must have paid their IBS annual
subscription for 2007 prior to the close of registration on
April 15th 2007.
The local organisers suggest that the best way to travel to Rolduc
is to first fly to either Brussels or Amsterdam, and then take a train
journey to Kerkrade. This train journey is about 3 hours from
Brussels and about 3 hours 15 minutes from Amsterdam. The local
organisers will organise a shuttle bus service from the railway
station at Kerkrade to the conference venue on both
Tuesday 8th May and Friday 11th May. Other
possible travel arrangements might include flights to Maastricht, or
Eurostar to Brussels followed by onward train to Kerkrade.
¤ ¤ ¤ ¤ ¤
¤ ¤ ¤ ¤ ¤ ¤ ¤
¤ ¤ ¤ Thanks to the continuing generosity of the Fisher Memorial
Trust, we have funds available to support four bursaries of £300
each to support the attendance of young biometricians (defined as
being aged 35 or within 10 years of completing their PhD) at the
First Channel Network Conference in Rolduc, Netherlands from 8-11 May
2007. To be eligible, applicants should be intending to present at
the conference (note that abstracts for further poster contributions
can still be submitted until March 25th – see above).
Preference will be given to members of the International Biometric
Society, and those who have not previously received a
Fisher Memorial Trust Bursary. Applications, including a breakdown of your costs for attending the
conference, the abstract for your contribution to the conference,
and an indication of your eligibility as a `young biometrician',
should be sent to the British and Irish Region Secretary,
Andrew Mead
(andrew.mead@warwick.ac.uk), to
reach him before the end of March. Successful applicants will be
informed by April 10th.
225 euros - for members of the IBS if registering prior to
March 25th 2007
275 euros - for members of the IBS if registering after
March 25th 2007
275 euros - for non-members if registering prior to
March 25th 2007
325 euros - for non-members if registering after
March 25th 2007
Accommodation
Accommodation can be booked directly with the Rolduc
Conference Centre through a link on the conference web-site.
Travel
Important Dates
March 25th
Early-bird registration deadline
April 15th
Final registration deadline
May 8th - 11th
First Channel Network Conference
Fisher Memorial
Trust Bursaries for Young Biometricians