| Join Now
Please find below the details of a fully-funded PhD position in Tours (France): "Estimating complex intervention effects as risk differences in cluster randomised trials".
If you wish to apply or if you would like more information, please contact Prof. Bruno Giraudeau bruno.giraudeau `at' univ-tours.fr
3-year PhD position in biostatistics
Estimating complex intervention effects as risk differences in cluster randomized trials
Cluster randomized trials (cRTs) are defined as randomized trials in which clusters of
individuals are randomized rather than individuals themselves. Such trials may be conducted
in clinical settings and thus randomize practices, health professionals, wards, hospitals, etc. or
non-clinical settings and randomize schools, families, residential areas, worksites, etc. The
most compelling reason for cluster randomization is that the intervention cannot be delivered
individually, for instance, for education interventions of healthcare practitioners, for most
health service interventions or for health promotion interventions such as education programs
using general media (i.e., TV, radio, etc.). As an example, the Printemps study assesses a
tailored promotion intervention of a mobile application and website that offer evidence-based
content for suicide prevention. Most of the interventions assessed in cRTs are said to be
complex interventions, which means that they have several components. As an example, an
intervention aimed at reducing the use of antipsychotic drugs in nursing homes could involve
a health professional education program, the implementation of a new prescription tool, and
the distribution of leaflets to staff.
Randomizing clusters leads to hierarchical data: participants are embedded in clusters,
which are randomly allocated to groups. As a consequence, there exists a correlation in data,
which are then classically analyzed using a modelling approach such as a mixed model or
generalized estimating equation. When the outcome is binary, a logit link function is
classically used, which leads to expressing the intervention effect as an odds ratio. However,
because expressing an intervention effect on a relative scale is associated with an overoptimistic
appraisal of the result, results should be reported both as absolute and relative
effect sizes. For binary data, this involves assessing a risk difference.
The objective of this PhD position is to identify the statistical approach with the best
statistical properties (Type I error rate, bias, coverage rate, power) to estimate an adjusted risk
difference from correlated clustered data, considering a two-parallel group cRT.
Assessing the statistical properties of the different identified methods will involve
simulation studies. Results will be illustrated by using real data from cluster trials. As
examples, the Ambroisie study assessed a strategy of gastric emptiness peri-extubation in
intensive care unit patients, the Pralimap study assessed interventions aimed at preventing
obesity in adolescents in grade 9 and 10, and the Apache 3 study compared two strategies to
motivate unscreened women to participate in a whole cervical-cancer screening program.
This PhD position corresponds to one of the work-packages of the ESCIENT project, funded
by the French National Agency of Research (ANR – AAPG 2021). The ESCIENT project
focuses on estimating a complex intervention effect when the outcome is binary:
Complex interventions include several components that may interact with one another.
In health service and public health research, they are often assessed with cRTs, which
randomize intact social units. In these trials, which population to be analyzed is a
challenging issue. The boundary between lack of compliance and adaptation of the
intervention to the context is tenuous; participants may lack compliance because of
their own will or because of external events; because the intervention may be adapted
to the context, an “as-treated” population is also of interest. We definitely need
guidelines on this issue. From a statistical viewpoint, such trials are also challenging.
An adjusted risk difference is the preferred way of expressing results, whereas
classical statistical analyses usually return odds ratios. We presently lack clear
recommendations on the optimal statistical method to be used. The ESCIENT project
aims to close these two gaps.
Two research units are involved in this project: the unit methodS in Patient-centered outcomes
and HEalth ResEarch (SPHERE, https://sphere-inserm.fr) involves experts of cRTs and
APEMAC (https://apemac.univ-lorraine.fr/) involves experts of complex interventions.
UMR INSERM 1246 – SPHERE
The PhD student will join the INSERM U1246 research unit. This unit, entitled SPHERE
(https://sphere-inserm.fr) is jointly accredited by the universities of Tours and Nantes as well
as INSERM. It is a multidisciplinary unit that aims to develop and validate methods that can
be used in clinical or epidemiological studies. Researchers work considering a
pluridisciplinary perspective involving biostatistics, public health, clinical disciplines
(addictology, dermatology, general practice, nephrology), pharmacology, health psychology,
and health economics. The director is Véronique Sébille (University of Nantes), and Bruno
Giraudeau (University of Tours) is the deputy director.
Bruno Giraudeau will supervise the PhD candidate. He is a professor of biostatistics
(https://orcid.org/0000-0003-3031-8258) and works both at the University and Hospital of
Tours. The PhD student will be located in Tours, and the doctoral school will be the Health,
Biological Sciences and Life Chemistry doctoral school.
Because this PhD position is part of a larger project, there will be interactions with other
- researchers from the APEMAC research unit involved in the ESCIENT project: Nelly
Agrinier, Laëtitia Minary and Joëlle Kivits (Nancy, France)
- members of the scientific committee who are experts in cRTs: Monica Taljaard
(Ottawa, Canada), Sandra Eldridge (London, UK), Elisabeth Turner (Duke, USA) and
Agnès Caille (Tours, France).
The net salary will be about 1700€/month. Of note, in Tours, a flat can be rented for 400€ to 500€/month.
Annual tuition fee
Completion of a MSc in biostatistics, medical statistics.
Advanced programming skills in the statistical software program R.
Living in Tours
With about 136 000 inhabitants (360 000 in the
conurbation), Tours is a human-scale city pleasant to live in.
It is a charming student town (30 000 students at the
university), benefiting from mild weather along the Loire
river. Near famous castles such as Chenonceau, Chambord,
Villandry or the Clos Luce where Da Vinci rests, Tours is
just a one-hour train ride from Paris.
•: INSERM U1246 - SPHERE
2ième étage du bâtiment tertiaire
CHRU de Tours
2 Bd Tonnellé
37044 Tours cedex 9
@: bruno.giraudeau `at' univ-tours.fr
Statistical Society of Australia (SSA)
PO Box 213
Belconnen ACT 2616 Australia
02 6251 3647www.statsoc.org.auABN 82 853 491 081
Please direct enquiries to:
Marie-Louise Rankin, Executive Officer
© 2019 Statistical Society of Australia (SSA). All Rights Reserved. | website login
Website by Converge Design