Explore chapters and articles related to this topic
Assessment of Moderator and Mediator Effects
Published in Gueorguieva Ralitza, Statistical Methods in Psychiatry and Related Fields, 2017
In order to deal with the interpretability issue and to formalize conditions under which causal inference can be performed, a few publications in the statistical literature have considered mediation analysis with time-varying exposures and repeated assessments of the mediator (VanderWeele, 2010;Bind et al., 2016). An approach that has gained popularity in the statistical and epidemiology literature to assess mediation effects when treatment is not randomized and there may be time-varying confounders and exposures is marginal structural models (MSM, see e.g., Moodie and Stephens (2011) for a non-technical description). These models are marginal because they assess population-averaged effects of the treatment and the mediator on the outcome, and structural because they focus on causal rather than associational effects. There are different methods for estimation: inverse probability weighting (Robins et al., 2000), g-estimation (van der Wal et al., 2009), or targeted maximum likelihood (Rosenblum and van der Laan, 2010). These methods have not yet gained popularity in the subject-matter literature.
Causal Inference for Observational Studies/Real-World Data
Published in Harry Yang, Binbing Yu, Real-World Evidence in Drug Development and Evaluation, 2021
From a practical perspective, the propensity score is not known in advance and researchers must estimate it. Whether using parametric or nonparametric methods, there is always a chance that the propensity score is not estimated correctly. This is less an issue for matching because the estimated propensity score itself does not go into the estimator. As long as the matching obtains desirable balance, it should be fine for the design purpose, but the situation is different for weighting, as the estimated propensity score is part of the estimator. A mis-specified propensity score model will introduce substantial bias. To overcome this, researchers have developed a so-called “doubly robust” estimation strategy, which combines the propensity score and a regression model to improve the performance (Bang and Robins 2005). A regression model on the outcome is introduced to guard against potential mis-specifications of the propensity score model. Such a model is known as a structural model to differentiate from the conventional regression model, because it models the potential outcomes rather than the observed ones. A common type of structural model is the marginal structural model, which focuses on the marginal causal effect (Robins et al. 2000). The advantage of the doubly robust estimation strategy is that it has two chances to get a consistent estimate of the causal effect: (1) when the outcome model is incorrectly specified but the propensity score model is correctly specified or (2) when the propensity score model is incorrectly specified but the outcome model is correctly specified. When both models are correct, it yields an efficient causal effect estimator (Tan 2007).
Effects of Erectile Dysfunction Drugs Use on T-Cells and Immune Markers on Men Who Have Sex with Men
Published in International Journal of Sexual Health, 2022
Jee Won Park, Onyebuchi A. Arah, Otoniel Martinez-Maza, Adrian S. Dobs, Ken S. Ho, Frank J. Palella, Eric C. Seaberg, Roger Detels
Despite the aforementioned limitations, our study has notable strengths. To our knowledge, this is the first study that has examined the prospective immunologic and inflammatory effects of ED drug use over time among MSMs. Most of the prior evidence has been from animal studies, and there has been a lack of studies evaluating the prospective use of ED drugs in humans and MSMs. Another strength of this study included using a well-studied, large cohort of MSM in the U.S. This study design allowed for the examination of the effects of ED drug use on immune cells and markers over a 10-year time period, and included socially sensitive data collected through a standardized method. In addition, both MWH and MWOH were included in the analyses. Finally, marginal structural models produced estimates (i.e., mean differences) with efficient standard errors and was able to deal with time-varying exposure and confounding (Daniel et al., 2013; Wang et al., 2017).
Hospitalization risk in bipolar disorder patients treated with lurasidone versus other atypical antipsychotics
Published in Current Medical Research and Opinion, 2019
Daisy Ng-Mak, Rachel Halpern, Krithika Rajagopalan, Antony Loebel
Observational studies in bipolar disorder are challenging due to the dynamic nature of bipolar disorder episodes. Over time treatment is often modified to match the current symptoms. In order to more accurately estimate the treatment effect on the outcome, the common practice of treatment switching needs to be taken into account. By using a marginal structural model that examined treatment in each month and adjusted for known and measured time-sensitive confounding variables, the estimation of causal effects is considered to be improved26.
Hospitalization risk among adults with bipolar I disorder treated with lurasidone versus other oral atypical antipsychotics: a retrospective analysis of Medicaid claims data
Published in Current Medical Research and Opinion, 2021
Xiaoli Niu, Phani Veeranki, Syvart Dennen, Carole Dembek, Kimberly Laubmeier, Yanmei Liu, G. Rhys Williams, Antony Loebel
To control for time-varying confounders (i.e. factors which affect both the outcomes of interest and current or future treatments) and treatment, marginal structural models (MSMs) were used to estimate the inpatient admissions rate and hospital LOS22. The MSM design controls for treatment switching, which is frequent in patients receiving antipsychotics and complicates the estimation of the association of treatments with outcomes when using an ITT (intent-to-treat) approach23. MSMs have been used in observational studies of mood disorders using healthcare administrative claims data19. The stabilized inverse probability of treatment weights (IPTW) was calculated for each month in the post-index period using predicted probabilities from multinomial logistic regressions for the seven treatment classifications (lurasidone as reference, aripiprazole, olanzapine, quetiapine, risperidone, no/minimal treatment and other treatment). All-cause and bipolar I disorder-related hospitalization rates were modeled with generalized linear models with a logit link and clustered by patient to estimate the adjusted odds ratios (aORs) and 95% confidence intervals (CIs). All-cause and bipolar I disorder-related hospital LOS were modeled with zero-inflated Poisson regression models to estimate the adjusted incidence rate ratios (aIRRs) and 95% CI. Time-invariant covariates included age, sex, race, the pre-index period CCI score, pre-index period comorbidities, the pre-index period dependent variable and index year. Time-varying covariates included the prior-period treatment category, the prior-period dependent variable, the prior-period office visits, the prior-period substance abuse indicators and a post-index month indicator. Additional details of the MSM estimation are available in the Supplementary materials. All models were assessed for goodness of fit, and no multiple testing adjustments were performed.