Stijn Vansteelandt

Odds Ratios for Mediation Analysis for a Dichotomous Outcome

For dichotomous outcomes, the authors discuss when the standard approaches to mediation analysis used in epidemiology and the social sciences are valid, and they provide alternative mediation analysis techniques when the standard approaches will not work. They extend definitions of controlled direct effects and natural direct and indirect effects from the risk difference scale to the odds ratio scale. A simple technique to estimate direct and indirect effect odds ratios by combining logistic and linear regressions is described that applies when the outcome is rare and the mediator continuous. Further discussion is given as to how this mediation analysis technique can be extended to settings in which data come from a case-control study design. For the standard mediation analysis techniques used in the epidemiologic and social science literatures to be valid, an assumption of no interaction between the effects of the exposure and the mediator on the outcome is needed. The approach presented here, however, will apply even when there are interactions between the effect of the exposure and the mediator on the outcome.

Attributable mortality of ventilator-associated pneumonia: a reappraisal using causal analysis.

RATIONALE Measuring the attributable mortality of ventilator-associated pneumonia (VAP) is challenging and prone to different forms of bias. Studies addressing this issue have produced variable and controversial results. OBJECTIVES We estimate the attributable mortality of VAP in a large multicenter cohort using statistical methods from the field of causal inference. METHODS Patients (n = 4,479) from the longitudinal prospective (1997-2008) French multicenter Outcomerea database were included if they stayed in the intensive care unit (ICU) for at least 2 days and received mechanical ventilation (MV) within 48 hours after ICU admission. A competing risk survival analysis, treating ICU discharge as a competing risk for ICU mortality, was conducted using a marginal structural modeling approach to adjust for time-varying confounding by disease severity. MEASUREMENTS AND MAIN RESULTS Six hundred eighty-five (15.3%) patients acquired at least one episode of VAP. We estimated that 4.4% (95% confidence interval, 1.6-7.0%) of the deaths in the ICU on Day 30 and 5.9% (95% confidence interval, 2.5-9.1%) on Day 60 are attributable to VAP. With an observed ICU mortality of 23.3% on Day 30 and 25.6% on Day 60, this corresponds to an ICU mortality attributable to VAP of about 1% on Day 30 and 1.5% on Day 60. CONCLUSIONS Our study on the attributable mortality of VAP is the first that simultaneously accounts for the time of acquiring VAP, informative loss to follow-up after ICU discharge, and the existence of complex feedback relations between VAP and the evolution of disease severity. In contrast to the majority of previous reports, we detected a relatively limited attributable ICU mortality of VAP.

A Simple Unified Approach for Estimating Natural Direct and Indirect Effects

An important problem within both epidemiology and many social sciences is to break down the effect of a given treatment into different causal pathways and to quantify the importance of each pathway. Formal mediation analysis based on counterfactuals is a key tool when addressing this problem. During the last decade, the theoretical framework for mediation analysis has been greatly extended to enable the use of arbitrary statistical models for outcome and mediator. However, the researcher attempting to use these techniques in practice will often find implementation a daunting task, as it tends to require special statistical programming. In this paper, the authors introduce a simple procedure based on marginal structural models that directly parameterize the natural direct and indirect effects of interest. It tends to produce more parsimonious results than current techniques, greatly simplifies testing for the presence of a direct or an indirect effect, and has the advantage that it can be conducted in standard software. However, its simplicity comes at the price of relying on correct specification of models for the distribution of mediator (and exposure) and accepting some loss of precision compared with more complex methods. Web Appendixes 1 and 2, which are posted on the Journals Web site (http://aje.oupjournals.org/), contain implementation examples in SAS software (SAS Institute, Inc., Cary, North Carolina) and R language (R Foundation for Statistical Computing, Vienna, Austria).

Estimating Direct Effects in Cohort and Case-Control Studies

Estimating the effect of an exposure on an outcome, other than through some given mediator, requires adjustment for all risk factors of the mediator that are also associated with the outcome. When these risk factors are themselves affected by the exposure, then standard regression methods do not apply. In this article, I review methods for accommodating this and discuss their limitations for estimating the controlled direct effect (ie, the exposure effect when controlling the mediator at a specified level uniformly in the population). In addition, I propose a powerful and easy-to-apply alternative that uses G-estimation in structural nested models to address these limitations both for cohort and case–control studies.

Mediation Analysis with Multiple Mediators

Abstract Recent advances in the causal inference literature on mediation have extended traditional approaches to direct and indirect effects to settings that allow for interactions and non-linearities. In this article, these approaches from causal inference are further extended to settings in which multiple mediators may be of interest. Two analytic approaches, one based on regression and one based on weighting are proposed to estimate the effect mediated through multiple mediators and the effects through other pathways. The approaches proposed here accommodate exposure–mediator interactions and, to a certain extent, mediator–mediator interactions as well. The methods handle binary or continuous mediators and binary, continuous or count outcomes. When the mediators affect one another, the strategy of trying to assess direct and indirect effects one mediator at a time will in general fail; the approach given in this article can still be used. A characterization is moreover given as to when the sum of the mediated effects for multiple mediators considered separately will be equal to the mediated effect of all of the mediators considered jointly. The approach proposed in this article is robust to unmeasured common causes of two or more mediators.

On model selection and model misspecification in causal inference

Standard variable selection procedures, primarily developed for the construction of outcome prediction models, are routinely applied when assessing exposure effects in observational studies. We argue that this tradition is sub-optimal and prone to yield bias in exposure effect estimators as well as their corresponding uncertainty estimators. We weigh the pros and cons of confounder-selection procedures and propose a procedure directly targeting the quality of the exposure effect estimator. We further demonstrate that certain strategies for inferring causal effects have the desirable features (a) of producing (approximately) valid confidence intervals, even when the confounder-selection process is ignored, and (b) of being robust against certain forms of misspecification of the association of confounders with both exposure and outcome.

Effect decomposition in the presence of an exposure-induced mediator-outcome confounder

Methods from causal mediation analysis have generalized the traditional approach to direct and indirect effects in the epidemiologic and social science literature by allowing for interaction and nonlinearities. However, the methods from the causal inference literature have themselves been subject to a major limitation, in that the so-called natural direct and indirect effects that are used are not identified from data whenever there is a mediator-outcome confounder that is also affected by the exposure. In this article, we describe three alternative approaches to effect decomposition that give quantities that can be interpreted as direct and indirect effects and that can be identified from data even in the presence of an exposure-induced mediator-outcome confounder. We describe a simple weighting-based estimation method for each of these three approaches, illustrated with data from perinatal epidemiology. The methods described here can shed insight into pathways and questions of mediation even when an exposure-induced mediator-outcome confounder is present.

Preterm birth in twins after subfertility treatment: population based cohort study

Abstract Objectives To assess gestational length and prevalence of preterm birth among medically and naturally conceived twins; to establish the role of zygosity and chorionicity in assessing gestational length in twins born after subfertility treatment. Design Population based cohort study. Setting Collaborative network of 19 maternity facilities in East Flanders, Belgium (East Flanders prospective twin survey). Participants 4368 twin pairs born between 1976 and 2002, including 2915 spontaneous twin pairs, 710 twin pairs born after ovarian stimulation, and 743 twin pairs born after in vitro fertilisation or intracytoplasmic sperm injection. Main outcome measures Gestational length and prevalence of preterm birth. Results Compared with naturally conceived twins, twins resulting from subfertility treatment had on average a slightly decreased gestational age at birth (mean difference 4.0 days, 95% confidence interval 2.7 to 5.2), corresponding to an odds ratio of 1.6 (1.4 to 1.8) for preterm birth, albeit confined to mild preterm birth (34-36 weeks). The adjusted odds ratios of preterm birth after subfertility treatment were 1.3 (1.1 to 1.5) when controlled for birth year, maternal age, and parity and 1.6 (1.3 to 1.8) with additional control for fetal sex, caesarean section, zygosity, and chorionicity. Although an increased risk of preterm birth was therefore seen among twins resulting from subfertility treatment, the risk was largely caused by a first birth effect among subfertile couples; conversely, the risk of prematurity was substantially levelled off by the protective effect of dizygotic twinning. Conclusions Twins resulting from subfertility treatment have an increased risk of preterm birth, but the risk is limited to mild preterm birth, primarily by virtue of dizygotic twinning.

Placental inflammation and perinatal transmission of HIV-1.

&NA; The effect of placental membrane inflammation on mother‐to‐child transmission (MTCT) of HIV‐1 is reported. Placentas from HIV‐1‐infected women were examined as part of a perinatal HIV‐1 project in Mombasa. Kenya. Polymerase chain reaction analysis was used to test for HIV‐1 in the infants at birth and at 6 weeks. The maternal HIV‐1 seroprevalence was 13.3% (298 of 2,235). The overall rate of MTCT of HIV‐1 was 25.4%; polymerase chain reaction analysis revealed that of the 201 infants 6.0% (12) were already HIV‐1‐positive at birth (intrauterine transmission) and 19.4% (39) were infected during the peripartum period or in early neonatal life (perinatal transmission). The prevalence of acute chorioamnionitis was 8.8%, that of deciduitis was 10.8%, and that of villitis was 1.6%. Acute chorioamnionitis was independently associated with peripartum HIV‐1 transmission but not with in utero MTCT (17.9% vs. 6.7%, respectively; adjusted odds ratio, 3.9; 95% confidence interval, 1.2‐12.5; p = .025). Other correlates of perinatal MTCT were presence of HIV in the genital tract and in the babys oral cavity and a high maternal viral load in peripheral blood. The adjusted population attributable fraction of 12.8% (95% confidence interval, 1.5%‐22.8%) indicated that approximately 3% of MTCT could be prevented if acute chorioamnionitis was eliminated. We suggest that further research on the role of antimicrobial treatment in the prevention of chorioamnionitis and the reduction of peripartum MTCT needs to be performed.

Make the most of your samples: Bayes factor estimators for high-dimensional models of sequence evolution

BackgroundAccurate model comparison requires extensive computation times, especially for parameter-rich models of sequence evolution. In the Bayesian framework, model selection is typically performed through the evaluation of a Bayes factor, the ratio of two marginal likelihoods (one for each model). Recently introduced techniques to estimate (log) marginal likelihoods, such as path sampling and stepping-stone sampling, offer increased accuracy over the traditional harmonic mean estimator at an increased computational cost. Most often, each model’s marginal likelihood will be estimated individually, which leads the resulting Bayes factor to suffer from errors associated with each of these independent estimation processes.ResultsWe here assess the original ‘model-switch’ path sampling approach for direct Bayes factor estimation in phylogenetics, as well as an extension that uses more samples, to construct a direct path between two competing models, thereby eliminating the need to calculate each model’s marginal likelihood independently. Further, we provide a competing Bayes factor estimator using an adaptation of the recently introduced stepping-stone sampling algorithm and set out to determine appropriate settings for accurately calculating such Bayes factors, with context-dependent evolutionary models as an example. While we show that modest efforts are required to roughly identify the increase in model fit, only drastically increased computation times ensure the accuracy needed to detect more subtle details of the evolutionary process.ConclusionsWe show that our adaptation of stepping-stone sampling for direct Bayes factor calculation outperforms the original path sampling approach as well as an extension that exploits more samples. Our proposed approach for Bayes factor estimation also has preferable statistical properties over the use of individual marginal likelihood estimates for both models under comparison. Assuming a sigmoid function to determine the path between two competing models, we provide evidence that a single well-chosen sigmoid shape value requires less computational efforts in order to approximate the true value of the (log) Bayes factor compared to the original approach. We show that the (log) Bayes factors calculated using path sampling and stepping-stone sampling differ drastically from those estimated using either of the harmonic mean estimators, supporting earlier claims that the latter systematically overestimate the performance of high-dimensional models, which we show can lead to erroneous conclusions. Based on our results, we argue that highly accurate estimation of differences in model fit for high-dimensional models requires much more computational effort than suggested in recent studies on marginal likelihood estimation.