Richard Doll Prize Lecturer George Davey Smith Focuses On The Role
Of Epidemiology In The Age Of Big Data
Is
there a role for epidemiology in the age of big data? That is the
question posed and answered in the affirmative by the University of
Bristol’s George Davey Smith, the recipient of this year’s
Richard Doll prize, in his lecture delivered by video to the
International Epidemiological Association (IEA) at its triennial
meeting in Japan last month.
Davey
Smith did not hold his audience in suspense for very long as he
confessed early in the lecture that there is an exciting future for
epidemiologists in the era of big data and it is important for
epidemiologists to enthusiastically embrace the opportunities
afforded.
He outlined his talk to cover five topics, namely:
·
Epidemiology and causation--role of genetics and biomarkers
·
Exposure
and outcome assessment
·
Levels
of causation
·
Limits
of prediction
·
A role
for epidemiology
Early Career
Davey Smith began by recounting his experiences in studying
cardiovascular disease in the 1980’s when observational studies
suggested that elevated levels of HDL cholesterol could be protective
against cardiovascular heart disease (CVD). He described a study
demonstrating the challenges in disentangling HDL and triglyceride
measurements in studying CVD. He called these and other CVD
measurement issues “intractable problems” which influenced him to
study other more epidemiologically tractable questions, for example
ones around HIV/AIDS and diarrhoeal disease in childhood. These
appeared less susceptible to measurement issues and the results had
more direct implications for public health.
New Approaches
Davey Smith recounted how during the 2,000’s the incorporation of data
on molecular genetic variation into observational epidemiological
studies - and in particular Mendelian randomization (MR) - could be
used to overcome some biases in studies and help strengthen causal
inference.
He
highlighted the benefits of triangulation or comparing different study
approaches to obtain more reliable or accurate answers to research
questions, including those around the role of elevated levels of HDL.
Large scale randomized controlled trials and MR studies converged on
providing strong evidence that modifying circulating HDL levels did
not improve cardiovascular outcomes. It is difficult to think of how
any other developments in epidemiological methodology had made a
serious contribution to understanding this crucial issue.
In
discussing other transformational changes that have taken place around
exposure and outcome assessment, Davey Smith used DNA methylation and
how it indexes exposure to smoking more accurately than self-reported
smoking data. It can even assess in utero exposures that took place
decades earlier. Other innovations he described include the use of
cameras which can be worn by infants to collect digital data on
parent-child interactions, something Davey Smith described as “ a
complete transformation” and one small indicator of the opportunities
that can be afforded by digital and big data.
Levels of Causation
In
discussing levels of causation, he reminded the audience that much of
the big data collected is at the individual level and of itself may
not be of particular relevance to
interpreting
the effects of broader, underlying social and economic influences on
levels and trends in populations. He warned epidemiologists that it is
important not to become distracted by the excitement around big data
from the main objective of epidemiology to impact population health by
creating population level interventions. While MR is a powerful tool,
it does not in itself help us develop interventions at the population
level, said Davey Smith.
Doll on Luck
He
quoted from Richard Doll to highlight the role of luck or chance in
determining which individuals actually come down with disease, and to
point out that the influence of stochastic factors limits the accuracy
of predictions that will be possible about individual health outcomes.
According to Doll, “Whether [any particular] exposed subject does or
does not develop a cancer is largely a matter of luck; bad luck if the
several necessary changes all occur in the same stem cell when there
are several thousand such cells at risk, good luck if they don’t.
Personally, I find that makes good sense, but many people apparently
do not.” Davey Smith called
this fact “good news”
for epidemiologists because it means that population
level interventions – which is what epidemiologists study – are
crucial to improving public health.
Guiding Lights for Epi
In
enthusiastically embracing big data, Davey Smith exhorted his audience
not to forget that it is the factors at the population level that
should be the focus for epidemiologists and that attention to basic
principles will pay off. He encouraged epidemiologists to embrace the
important and exciting role of grasping the opportunities provided by
big data, whilst not being distracted from the major task of the
discipline.
The
lecture is available at:
https://tinyurl.com/y7c6shoh
with a 2 minute animation outlining Mendelian randomization
that it contains also available at:
https://tinyurl.com/yba55mag
■
|