How to use marginal structural models in randomized trials to estimate the natural direct and indirect effects of therapies mediated by causal intermediates
Restricted accessResearch articleFirst published online June, 2011
How to use marginal structural models in randomized trials to estimate the natural direct and indirect effects of therapies mediated by causal intermediates
Background Although intention-to-treat analysis is a standard approach, additional
supplemental analyses are often required to evaluate the biological relationship
among interventions, intermediates, and outcomes. Therefore, we need to evaluate
whether the effect of an intervention on a particular outcome is mediated by a
hypothesized intermediate variable.
Purpose To evaluate the size of the direct effect in the total effect, we applied the
marginal structural model to estimate the average natural direct and indirect effects
in a large-scale randomized controlled trial (RCT).
Method The average natural direct effect is defined as the difference in the
probability of a counterfactual outcome between the experimental and control arms,
with the intermediate set to what it would have been, had the intervention been a
control treatment. We considered two marginal structural models to estimate the
average natural direct and indirect effects introduced by VanderWeele
(Epidemiology 2009) and applied them in a large-scale RCT – the
Candesartan Antihypertensive Survival Evaluation in Japan (CASE-J trial) – that
compared angiotensin receptor blockers and calcium-channel blockers in high-risk
hypertensive patients.
Results There were no strong blood pressure-independent or dependent effects;
however, a systolic blood pressure reduction of about 1.9 mmHg suppressed all events.
Compared to the blood pressure-independent effects of calcium channel blockers, those
of angiotensin receptor blockers contributed positively to cardiovascular and cardiac
events, but negatively to cerebrovascular events.
Limitations There is a particular condition for estimating the average natural direct
effect. It is impossible to check whether this condition is satisfied with the
available data.
Conclusion We estimated the average natural direct and indirect effects through the
achieved systolic blood pressure in the CASE-J trial. This first application of
estimating the average natural effects in an RCT can be useful for obtaining an
in-depth understanding of the results and further development of similar
interventions.
Get full access to this article
View all access options for this article.
References
1.
ICH harmonized tripartite guideline. Statistical principles for
clinical trials. International conference on harmonization of technical
requirements for registration of pharmaceuticals for human use,
1998.
2.
FreedmanLSGraubardBISchatzkinA. Statistical validation of intermediate endpoints for
chronic diseases. Stat Med1992; 11:
167–78.
3.
RobinsJM. The control of confounding by intermediate
variables. Stat Med1989; 8:
679–701.
4.
RosenbaumPR. The consequences of adjustment for a concomitant
variable that has been affected by the treatment. J R
Statist Soc A1984; 147:
656–66.
5.
KaufmanJSMacLehoseRFKaufmanS. A further critique of the analytic strategy of
adjusting for covariates to identify biologic mediation.
Epidemiol Perspect Innov2004; 1: 4–4.
6.
RobinsJMGreenlandS. Identifiability and exchangeability for direct and
indirect effects. Epidemiology1992; 3:
143–55.
7.
Pearl J. Direct and indirect effects. In: Proceedings of
Seventeenth Conference on Uncertainty in Artificial Intelligence,
Morgan Kaufman, San Francisco, 2000.
8.
HafemanDMSchwartzS. Opening the black box: a motivation for the
assessment of mediation. Int J Epidemiol2009; 38:
838–45.
9.
PetersenMSinsiSEvan der LaanMJ. Estimation of direct effect.
Epidemiology2006; 17:
276–84.
10.
VanderWeeleTJ. Marginal structural models for the estimation of
direct and indirect effects. Epidemiology2009; 20:
18–26.
11.
RosenblumMJewellNvan der LaanM. Analysing direct effects in randomized trials with
secondary interventions: an application to human immunodeficiency virus prevention
trials. J R Statist Soc A2009; 172:
443–65.
12.
GoetgelukSVansteelandtSGoetghebeurE. Estimation of controlled direct
effects. J R Statist Soc B2008; 70:
1049–66.
13.
van der LaanMJPetersenML. Direct effect models. Int J
Biostat2008; 4: Article 23–Article
23.
14.
OgiharaTNakaoKFukuiT. Effects of candesartan compared with amlodipine in
hypertensive patients with high cardiovascular risks: candesartan antihypertensive
survival evaluation in Japan trial. Hypertension2008; 51:
393–98.
15.
KaufmanSKaufmanJSMacLehoseRF. Improved estimation of controlled direct effects in
the presence of unmeasured confounding of intermediate variables.
Stat Med2005; 24:
1683–702.
16.
HollandP. Statistics and causal inference (with discussion and
rejoinder). J Am Statist Assoc1986; 81:
945–70.
17.
VanderWeeleTJ. Concerning the consistency assumption in causal
inference. Epidemiology2009; 20:
880–83.
18.
RobinsJM. Semantics of causal DAG models and the identification
of direct and indirect effects. In: GreenPHjortNLRichardsonS. (eds). Highly Structured Stochastic SystemsOxford University Press, New
York, 200370–81.
19.
RobinsJMHernanMABrumbackB. Marginal structural models and causal inference in
epidemiology. Epidemiology2000; 11:
550–60.
20.
SatoTMatsuyamaY. Marginal structural models as a tool for
standardization. Epidemiology2003; 14:
680–86.
21.
FukuiTRahmanMHayashiK. CASE-J study group.
Candesartan antihypertensive survival evaluation in Japan (CASE-J)
trial of cardiovascular events in high-risk hypertensive patients: rationale,
design, and methods. Hypertens Res2003; 26:
979–90.
22.
Japanese Society of Hypertension Guidelines Subcommittee
for the Management of Hypertension. Guidelines for the
management of hypertension for general practitioners.
Hypertens Res2001; 24:
613–34.
23.
HernanMABrumbackBRobinsJM. Marginal structural models to estimate the causal
effect of zidovudine on the survival of HIV-positive men.
Epidemiology2000; 11:
561–70.
24.
HernanMABrumbackBARobinsJM. Estimating the causal effect of zidovudine on CD4
count with a marginal structural model for repeated measures.
Stat Med2002; 21:
1689–709.
25.
WestfallPHYoungSS. Resampling-based Multiple TestingJohn Wiley and Sons, New
York, 1993.
26.
Blood Pressure Lowering Treatments Trialists’
Collaboration. Blood pressure-dependent and independent
effects of agents that inhibit the renin–angiotensin system.
J Hypertens2007; 25:
951–58.
27.
TaylorJMGWangYThiebautR. Counterfactual links to the proportion of treatment
effect explained by a surrogate marker.
Biometrics2005; 61:
1102–11.
28.
BurzykowskiTMolenberghsGBuyseM. The Evaluation of Surrogate EndpointsSpringer, New York,
2005.
29.
VanderWeeleTJ. Bias formulas for sensitivity analysis for direct and
indirect effect. Epidemiology2010; 21:
540–51.
30.
GreenlandS. Basic methods for sensitivity analysis of
biases. Int J Epidemiol1996; 25:
1107–16.
31.
BrumbackBAHernanMAHaneuseSRobinsJM. Sensitivity analyses for unmeasured confounding
assuming a marginal structural model for repeated measures.
Stat Med2004; 23:
749–67.
32.
KlungsøyrOSextonJSandangerINygardJF. Sensitivity analysis for unmeasured confounding in a
marginal structural Cox proportional hazards model.
Lifetime Data Anal2009; 15:
278–94.
33.
KaufmanSKaufmanJSMacLehoseRF. Improved estimation of controlled direct effects in
the presence of unmeasured confounding of intermediate variables.
Stat Med2005; 24:
1683–702.
34.
CaiZKurokiMPearlJTianJ. Bounds on direct effects in the presence of
confounded intermediate variables. Biometrics2008; 64:
695–701.
35.
SjolanderA. Bounds on natural direct effects in the presence of
confounded intermediate variables. Stat Med2009; 28:
558–71.
