SECOND ANNOUNCEMENT - includes job description

Imperial College of  Science, Technology and Medicine
Division of Primary Care and Population Health Sciences


You will have a PhD in statistics, or epidemiology, and relevant research
experience to support two studies in the Department of Epidemiology and
Public Health: ‘Case-control  Methodology and Spatial Epidemiology’
(ESRC-funded), and Genetic Epidemiology of Severe Bacterial Infections
(Wellcome Trust-funded). The post is for two years. Starting salary will be
£19,704, inclusive of London Allowance.

For an informal discussion, please contact Dr Jon Wakefield: 0171 594 3336
([log in to unmask])

To apply, please send a letter of application, CV, and the names of two
referees to
Simon Sheffield (fax: 0171 - 402 2150, email: [log in to unmask])
Department of Epidemiology and Public Health, Imperial College School of
Medicine, Norfolk Place, London W2 1PG. Closing date 26 January 1999

The College is striving towards Equal Opportunities.


The post-holder will have a PhD in statistics and will be employed,
primarily, to work on an ESRC-funded project, case-control methodology and
spatial epidemiology, under the supervision of Jon Wakefield and Paul
Elliott at the Department of Epidemiology and Public Health. You will also
assist with the statistical analysis of a Wellcome Trust-funded study on the
genetic epidemiology of severe bacterial infections; role of mannose binding
protein and tumour necrosis factor gene polymorphisms, supervised by Dr
Marjo-Riitta Jarvelin and Professor Paul Elliott.


This grant was obtained with Mandy Chetwynd and Peter Diggle at the
Department of Statistics at Lancaster University, and there will be
collaboration with this department. The project description follows, and it
is envisaged that the successful applicant will concentrate on the
errors-in-variables, Bayesian and socio-economic aspects of the project.

There is now considerable interest in the geographic variation of health
outcomes. The analysis of
geographically-indexed health data aims to quantify spatial variability in
risk and explain the variation in terms of socio-economic effects such as
deprivation and environmental factors such as pollution sources.

The analysis of routine health data can provide vital clues to aetiology but
such data are generally, and
in particular in the UK, available only at an aggregated level and the
analysis and interpretation of such data are fraught with well-documented

The case-control study, difficulties with bias notwithstanding, remains the
‘gold standard’ for investigating the relationship between health,
socio-economic effects and environmental factors.

In practice, case-controls studies often use stratified, or matched,
controls in order to provide a more efficient analysis by adjusting for
known risk factors which are not of direct interest. It is well known in
classical epidemiology that the resultant analysis has to acknowledge the
stratified/matched nature of the data but until now this has been ignored in
spatial epidemiology. For example describe a study investigating the
relationship between air pollution and lung cancer in Italy. They used the
point-process framework of but did not adjust for matching, thus producing
an incorrect analysis.

In general direct measures of environmental and socio-economic variables are
unavailable on study individuals and instead surrogates are used. Particular
examples include measures of deprivation and
environmental pollutants. Deprivation is strongly associated with many
diseases. It is also well known that socio-economic effects are
geographically related to environmental factors. For example there is a
tendency for the most deprived to live in close proximity to sources of
environmental pollution. Consequently, it is vital to provide methods for
subtle adjustment for this socio-economic confounding. Over-correcting for
these factors can mask a genuine environmental effect.

The data on these factors are available at an aggregated level, for example
based on census questions,
and so individuals may be assigned the value of the area within which they
are located. This must be acknowledged in the analysis, however. When
analysing case-control data in environmental contexts it is not possible to
obtain a measure of the primary exposure of interest (for example life-long
dose due to a particular pollution source) and so instead a surrogate is
frequently used. This surrogate may be derived from local pollution monitors
or may be the distance to a putative source. In general the use of surrogate
variables leads to attenuated estimates of the risk/exposure relationship
and only recently have errors-in-variables models been used in environmental

In this project we will develop methodology to carry out stratified/matched
spatial case-control studies whilst specifically acknowledging the
errors-in-variables nature of explanatory variables.


Susceptibility to infection is determined by host genetic factors as well as
overcrowding and deprivation. In a population-based nested case control
study, this project will define the roles of polymorphic alleles of two
candidate genes, mannose binding protein and tumour necrosis factor, in
susceptibility to, and outcome of, severe and recurrent infections. The
subjects will be those who developed severe or recurrent infections in a
birth cohort of 12,000 from Northern Finland who have been followed to age

The genotyping data will be held by the Department of Epidemiology and
Public Health which will link them to clinical data. The hospital records of
the cases will be reviewed, and details of number of infections, length of
hospital stay, severity based on clinical diagnosis, and antibiotic therapy
will be recorded. The allele frequencies will be determined in each of the
infection groups and controls using the Hardy Weinberg equation. The
statistical analysis will involve basic analysis of 2x2 tables (chi square,
odds ratios and 95% confidence limits) for matched data and multiple
logistic regression to adjust for potentially confounding variables such as
birth order and socio-economic status.


The Department of Epidemiology and Public Health has around fifty staff
working under the departmental head, Professor Paul Elliott, and has an
excellent track record of securing high levels of programme and research
grant funding across its three main areas of work. These are Environmental
Epidemiology & Small-Area Health Statistics (including the government-funded
Small Area Health Statistics Unit), Reproductive Epidemiology & Sexual
Health, and Cardiovascular Epidemiology.

The Department is located in refurbished offices in South Wharf Road on the
St. Mary’s (Paddington) site, and has a computing facility networked to
Imperial College.


· ability to prioritise jobs and organise work effectively;
· a proactive, self-starter with the ability to learn new skills quickly;
· able to work with a minimum of supervision;
· good, proven inter-personal communications skills
· able to work effectively with staff of all grades
· appropriate computing/technical skills


· to provide statistical support to the above research projects, and other
work, including small-area health statistics studies, in consultation with
the Head of Department
· to contribute to undergraduate and/or postgraduate teaching, as required
by the Head of Department
· statistical consultancy in consultation with the Head of Department

Other Tasks: as required by Professor Elliott, Head, Department of
Epidemiology and Public Health


The successful applicant will be accountable to Professor Elliott, with
day-to-day responsibility to lead investigators on the various research

Terms and Conditions: the appointment will be on the RA1A scale, point 6,
which is currently £19,704, inclusive of London Allowance.  The appointment
will be for two years in the first instance, with a probationary review at
six months. Closing date: 26 January 1999.