In April 2018 I retired from the Luxembourg Institute of Health, a publicly funded health care research institute in Luxembourg. I no longer take on postgraduate students. The list of topics here are thus mainly of historical rather than potential future interest
Basically I
am interested in any aspect of statistics that has to do with the
pharmaceutical industry and drug development. There is a list of my
publications: here However, and my blogs may also be of interest. My
interests can be organised under the following headings.
I make some
brief comments about these interests below.
A
particular emphasis of this work has been design and analysis of cross-over
trials and I have a book on the subject with Wiley (first edition 1993, second
edition 2002) Cross-over
Trials in Clinical Research. My collaborators in this area have
included Andy Grieve (several times), and my PhD students Dimitris Lambrou,
Sally Lee and Boikanyo Makubate
as well as Farkad Ezzet, Pina D'Angelo, Diane Potvin, Denise Till, Juergen Lillienthal, Francesco Patalano
and Hartwig Hildebrand. Bioequivalence is a
particular field of application for cross-over trials that has interesting
philosophical implications and is also something I have written on. In
particular I believe that more pharmacokinetic and pharmacodynamic
insights should be employed in designing and analysing
cross-over trials. I also believe that cross-over trials have a very important
role to play in personalized medicine.
However, I have also been
interested in parallel group trials, and in particular anything to do with
designing and interpreting multi-centre trials.
(Actually that is more or less all parallel group trials but what I mean
by this is that it is specifically the multi-centre
aspect of such trials that interests me.) My PhD students Andy Garrett and
Steven Julious have worked in problems to do with
parallel group studies, the former with matters to do with subgroup analysis
and the latter with sample size determination when nuisance parameters are
unknown. See also my book Statistical
Issues in Drug Development (Wiley, 1997, second edition 2007, third edition due 2021) for
examples of some of the interesting problems that can arise.
Most of my work in the area
of clinical trials is readily accessible but one of my favourite
papers is in a Liber Amicorum for Roel van Strik (Erasmus
University, 1996) and is hard to get hold of and is therefore included here on
this website. The AB.BA Cross-over. How to perform the
two-stage analysis if you can't be persuaded that you shouldn't.
I am also interested in the
use of prognostic information in analysing clinical
trials and have had separate collaborations with Emmanual
Lesaffre (
Measurement is also a theme
that interests me. In my opinion many clinical measurements used are wasteful
of information and not a few are potentially very misleading. I have been
collaborating in this area with Steven Julious and
more recently also with Katie Rolfe.
I am
particularly interested in anything to do with P-values or with parallels with
and differences between Bayesian and frequentist modes of inference. (See, for
example, You May
Believe You are a Bayesian but You are Probably Wrong.)
I also am interested in issues to do with randomisation
and additivity and also causal inference (although I
have not produced much in the latter area yet - but I have ambitions!) and have
done some work on multiplicity with Frank Bretz.
My research
in this field has been mainly on the use of placebos and the applications of Rawlsian theories of justice to the conduct of clinical
trials. I am also interested in the issue of honest reporting of scientific
findings both in and out of the pharmaceutical industry.
My
interests here are in weighting to allow for imprecisely estimated variances
and also in interpretation of effects in the presence of varying background
risk as well as comparison of Bayesian and frequentist approaches.
A
particular concern is that current approaches to reporting safety data,
including in monitoring clinical trials, are very destructive of information. I
have participated in a number of data safety monitoring boards and am
interested in principle in trying to develop methods for better monitoring.
During the
last two of the eight years (1987-1995) I worked for CIBA-Geigy,
I was involved, amongst other matters, in looking at the question of how one
decides which drugs to develop. My interest in decision analysis grew out of
that. I have some papers on the subject of portfolio management. I have
collaborated with Carl-Fredrik Burman (Astra-Zeneca)
and Andy Grieve (Pfizer) on applications of decision analysis to drug
development. Again this is a topic that has an important role to play in
personalized medicine.
While
researching the first edition of my cross-over trials book I became intrigued
by the story of Cushny and Peebles and the trial they
organised at Kalamazoo, the data from which were
(incorrectly) reported by Student in his famous paper in Biometrika.
Bill Richardson and I published a paper in Statistics in Medicine (1994)
on the subject. I have remained interested in the history of statistics and
have also written a paper about RA Fisher's only involvement in a clinical
trial. The history of statistics is also a feature of my book with
Modelling
I was involved in the
multi-disciplinary EPSRC-funded SCAM project. More details are given here The SCAM project:
Simplicity, Complexity and Modelling. Together
with other members of the team, Peter Challenor,
Andrew Cliffe, Mike Christie, Philip Dawid, Suraje Dessai,
Jim Hall, Zoran Kapelan and
Jeremy Oakley, I have written a book for Wiley. Simplicity,
Complexity and Modelling
More recently I have
collaborated with Nick Holford and Hans Hockey on modeling observations below
the limit of quantitation (specifically in the context of pharmacokinetics).
Rob Donald did a PhD with me as one of his supervisors in real-time modeling in the
context of intensive care. I would like to claim credit for this but in fact
the idea is all his own!
In my
opinion this whole field is in a terrible mess. Some excellent work has been
done but it is lost in the hype and the noise. Contrary to popular belief, for
most diseases, we do not know what percentage of patients respond because the
work on isolating the components of variation, between patient, within patient
and patient by treatment interaction has not been done. Progress in this area will
depend on a synthesis of research from laboratory to patient. Unfortunately at
the moment it is being driven by the former. This is likely to continue to be the
focus of my research over the next few years.
Comets, Contagion and Contingency
A paper criticising an aspect of Hoyle and Wickramasinghe's cometary theory
and presented to the International Society for Clinical Biostatistics in
An incomplete blocks cross-over trial in asthma. A paper published as, ������ Senn, SJ, Lillienthal, J, Patalano, F, Till, MD, An incomplete blocks cross-over in asthma: a case study in collaboration, in Cross-over Clinical Trials, Vollmar, J., and Hothorn, L. A., Eds., Fischer, Stuttgart;1997 pp. 3-26, but available here in pre-print form.
These are datasets to
accompany various papers I have written.
1. Programs to
accompany Cross-over Trials in Clinical
Research .2, SAS� macros for meta-analysis
(First author Jim Weir and important input from Tsushung
Hua, Connie Berlin, Mike Branson, Ekkehard
Glimm) 3. Programs for analysing cross-over trials in infertility (Mainly written
by Boikanyo Makubate.)
Previously, I had
suggestions for topics in the following areas. P indicates practical and T theoretical and the
number of stars indicates the degree of each. A D indicates a topic that
is likely to be difficult.
� Decision analysis and sequential bioequivalence P* T*** D (Taken by Odile
Coudert, MSc awarded 2006)
� Cross-over trials in infertility P* T*** (Taken by Boikanyo Makubate, PhD awarded 2009)
� Translation into SAS� and R of an existing Mathcad�� computer tool for sample-size determination
for cross-over trials (MSc only) P*** T*� (Taken
by Michael Tracy, MSc� awarded 2009)
� Power comparison
of Prescott's test and the Mainland-Gart test for cross-over trials with binary
outcomes P* T***
� Random effects for
variance estimates for meta-analysis P*T*** (Taken by Giorgio DiGessa, MSc awarded 2008)
� Realistic priors for related
parameters P*T***D
� Quality of meta-analyses, an
empirical investigation� P**** (Taken
by Edith Eze, MSc awarded 2009)
� Implications of size (of trial)
dependent effects for meta-analysis P* T***
�
Random
effects variances and effect sizes in meta-analysis an empirical investigation
P**T* (Taken by Nicola Greenlaw,
MSc awarded 2010)
� Discrepancies
between meta-analysis and large randomized clinical trials P***T*
�
Covariate
adjustment for ordered categories P* T***
� Comparison of propensity score and
analysis of covariance P** T**
� Problems with propensity based multiple imputation P** T*** (Taken by Jim Weir, MSc awarded
2006)
� Decision analytic approaches to
using biomarkers and surrogate endpoints in clinical trials T**** D
� Two-tailed tests for discrete data
T****
�
Predicting patient
recruitment P*** T* (Taken by Andisheh Bakhshi. MSc awarded 2012)
� Analysis of data on spontaneous reports of adverse events
associated with drugs P**T**D (Taken by Emmanuel Baah,
PhD awarded 2014)
� Decision analysis and implication
for ethics of clinical trials T**** D
�
Minimisation
and randomisation in clinical trials T**** D
�
Equivalence
of Inhaled Formulations.htm� P**T**D
�
Registering
pharmaceuticals for treating children.htm P*T***D
�
Statistical
issues in personalising medicine.htm T****D
I maintain a declaration of all* my interests in order that all those who find this a useful way to judge work may judge my work accordingly.
*That is to say all that I have not missed due to oversight
Back to Stephen Senn's
homepage
This page last updated 23
February 2021.