BEGIN:VCALENDAR
VERSION:2.0
PRODID:-//talks.cam.ac.uk//v3//EN
BEGIN:VTIMEZONE
TZID:Europe/London
BEGIN:DAYLIGHT
TZOFFSETFROM:+0000
TZOFFSETTO:+0100
TZNAME:BST
DTSTART:19700329T010000
RRULE:FREQ=YEARLY;BYMONTH=3;BYDAY=-1SU
END:DAYLIGHT
BEGIN:STANDARD
TZOFFSETFROM:+0100
TZOFFSETTO:+0000
TZNAME:GMT
DTSTART:19701025T020000
RRULE:FREQ=YEARLY;BYMONTH=10;BYDAY=-1SU
END:STANDARD
END:VTIMEZONE
BEGIN:VEVENT
CATEGORIES:Statistics
SUMMARY:The concept of separable effects for causal mediat
ion and competing risks analyses - Vanessa Didelez
\, Leibniz Institute for Prevention Research and E
pidemiology - BIPS\, Bremen\, Germany
DTSTART;TZID=Europe/London:20200221T140000
DTEND;TZID=Europe/London:20200221T150000
UID:TALK135964AThttp://talks.cam.ac.uk
URL:http://talks.cam.ac.uk/talk/index/135964
DESCRIPTION:In causal mediation analysis\, we are interested i
n understanding different mechanisms (causal pathw
ays) of a treatment or exposure affecting some out
comes. Often this is formalised in terms of (in)di
rect causal effects - popular notions of these are
based on so-called “nested counterfactuals”. Thes
e concepts however run into difficulties of interp
retation in the particular context of survival ana
lyses. \n\nI will discuss the problem and propose
an alternative approach that does not suffer from
such shortcomings [1]: this novel approach follow
s Robins and Richardson [2]\, where mechanisms nee
d to be specified allowing a separation into the d
ifferent treatment paths\, formalized using an aug
mented directed acyclic graph (DAG). It can be sho
wn that under specific assumptions regarding the s
eparability\, identification of such alternative m
ediated effects is possible\, resulting in the fam
iliar mediation formula. In continuous time\, it c
an further be shown that for the particular case o
f combining a linear model for the mediator with a
n additive hazard model\, the familiar “path-traci
ng” formula can be recovered [3].\n\nFor illustrat
ion\, this is applied to an example of mediated ef
fects of a blood-pressure treatment on time to kid
ney failure [3]. We investigate intensive versus s
tandard blood-pressure treatment and find that the
re is little\, and not much time-varying\, indirec
t effect via diastolic blood pressure on kidney fa
ilure. Hence\, other ways of preventing this side
effect of intensive blood-pressure treatment might
be worth investigated.\n\nThe proposed new approa
ch solves a crucial conceptual problem of mediatio
n analysis with a survival outcome and can be exte
nded to competing risks [4]. It is founded in deci
sion theory\, avoids genuine counterfactual assump
tions and constitutes an interesting alternative t
o the popular structural equation models.\n\nRefer
ences\n\n[1] Didelez. Defining causal mediation wi
th a longitudinal mediator and a survival outcome.
Lifetime Data Analysis\, DOI: 10.1007/s10985-018-
9449-0\, 2018.\n\n[2] Robins\, Richardson. Alterna
tive graphical causal models and the identificatio
n of direct effects. In: Causality and psychopatho
logy: Finding the determinants of disorders and th
eir cures\, pages 103-158\, 2011.\n\n[3] Aalen\, S
tensrud\, Didelez\, Daniel\, Roysland\, Strohmaier
. Time-dependent mediators in survival analysis: M
odelling direct and indirect effects with the addi
tive hazards model. Biometrical Journal. 2019\; (E
pub 2019 Feb 19).\n\n[4] Stensrud\, Young\, Didele
z\, Robins\, Hernán. Separable effects for causal
inference in the presence of competing risks. To a
ppear in JASA.
LOCATION:MR12
CONTACT:Dr Sergio Bacallado
END:VEVENT
END:VCALENDAR