| |
|
|
| |
[< back to Course Program] |
|
| |
|
|
| |
Session |
Day/Time
|
Session Topic and Speaker |
| K |
Sunday
11:15-12:45 |
Overview of New Approaches to Causal Models (Causal Pathways DAGs, Mathematical Models, FPDR Bayesian Approaches)
• Maya Petersen
Division of Biostatistics, School of Public Health, University of San Francisco, San Francisco, CA, United States
|
Abstract: |
Causal Diagrams in Epidemiology: an Application to the Direct Effect of SNPs
The effects of longitudinal exposures can be confounded by factors that are themselves affected by past exposure. Inverse probability weighting is a well-recognized method that allows for control of such time-dependent confounding, as well as providing a means to adjust for informative censoring. Although perhaps a less well-recognized application, inverse probability weighting can also be used to estimate the effects of dynamic treatment regimes, or treatment strategies that respond to changes in patient characteristics over time. This talk reviews alternative applications of inverse probability weighting, with an emphasis on understanding the types of questions this class of methods can be used to answer and on practical issues that arise in the course of their implementation. The talk demonstrates how inverse probability weighting can be used to provide insight into the antiretroviral treatment of HIV-infected subjects. The following research questions are addressed: 1) How rapidly do CD4+ T lymphocyte counts decline among subjects who remain on a virologically failing antiretroviral regimen? 2) How does delayed regimen modification following virologic failure of antiretroviral therapy affect mortality? 3) To what extent do mortality rates differ depending on the criteria used to detect failure and initiate regimen modification? These applications are illustrated using data from two observational clinical cohorts (from Johns Hopkins and the University of North Carolina, Chapel Hill). Implications of results for strategies to expand access to antiretroviral therapy in resource-limited settings are discussed, with a particular focus on the recent controversy regarding the need for viral load monitoring to detect treatment failure. |
Recommended Literature: |
- Robins JM, Greenland S. Identifiability and exchangeability for direct
and indirect effects. Epidemiology. 1992;3:143–155.
- Pearl J. Direct and indirect effects. In: Proceedings of the Seventeenth
Conference on Uncertainty in Artificial Intelligence. San Francisco:
Morgan Kaufmann; 2001:411– 420.
- Robins JM. Semantics of causal DAG models and the identification of
direct and indirect effects. In: Hjort N, Green P, Richardson S, eds.
Highly Structured Stochastic Systems. Oxford: Oxford University Press;
2003:70–81.
- Petersen ML, Sinisi SE, van der Laan MJ. Estimation of direct causal effects. Epidemiology 2006; 17(30) 276-84
- Cole SR, Hernan MA. Fallibility in estimating direct effects. Epidemiology.
2002;31:163–165.
- Poole C, Kaufman JS. What does standard adjustment for down- stream
mediators tell us about social effect pathways. Am J Epidemiol. 2000;
151:s52.
- Joffe MM, Colditz GA. Restriction as a method for reducing bias in the
estimation of direct effects. Stat Med. 1998;17:2233–2249.
- Kaufman S, Kaufman JS, Maclehose RF, et al. Improved estimation of
controlled direct effects in the presence of unmeasured confounding of
intermediate variables. Stat Med. 2005;24:1683–1702.
- Taylor JM, Wang Y, Thieaut R. Counterfactual links to the proportion of treatment effect explained by a surrogate marker. Biometrics 2005; 61(4): 1102-11
- van der Laan MJ, Petersen ML. Direct Effect Models. To be published in International Journal of Biostatstics. Technical Report available at: http://works.bepress.com/maya_petersen/3
|
Biography: |
|
Maya Petersen is an Assistant Professor of Biostatistics and Epidemiology at the University of California, Berkeley. She is currently on leave from her faculty appointment to complete an M.D. at the University of California, San Francisco. Maya received a Ph.D. in Biostatistics from the Berkeley School of Public Health, where her doctoral work was funded by the Howard Hughes Medical Institute and was honored by the Evelyn Fix prize. Maya’s research interests include the development and application of new causal inference methods, the treatment of HIV resistant to antiretroviral drugs, and the use of antiretroviral therapy in resource-limited settings. |
|
|
| |
|
|
|
