Just a moment, the page is loading...

Statistical Methods for Causal Inference Adjusting for Adherence

Statistical Methods for Causal Inference Adjusting for Adherence

Daniel Heitjan

Southern Methodist University Department of Statistics University of Texas Southwestern Medical Center Department of Clinical Sciences

None. The data will be used for statistical methods research for a PhD dissertation. The student, Ms. Shuang Li, is a doctoral student in Statistics at Southern Methodist University, and is supported by teaching assistantships. The PI, Dr. Daniel Heitjan, is a Professor of Statistical Science at SMU and Professor of Clinical Sciences at UT Southwestern Medical Center. He receives salary support from those institutions to support his graduate teaching and research activities.


13 April 2018

In clinical drug trials for chronic diseases we commonly observe that many enrollees adhere imperfectly to their assigned treatment. When this occurs, it is difficult to extract causal inferences from the data, as subjects have essentially selected their own treatment, “breaking” the randomization. In the past 25 years, statisticians have developed an array of methods, under the rubric of “causal inference”, for the analysis of trials subject to nonadherence (more commonly denoted “noncompliance” in the statistical literature).
To develop, test, and illustrate such methods, it is critical to have access to important, interesting, contemporary, real-world data sets. Key publications to date have used data from the LRC-CPPT trial on cholestyramine (Efron and Feldman 1991; Goetghebeur and Molenberghs 1996; Jin and Rubin 2008). This is a venerable data set, but the fact that it includes men only limits its value for our aim #3, which is to develop methods for extrapolating causal inferences from a trial population to the general population. Moreover the medication evaluated in that trial is no longer in wide use. A second key data set is from a randomized encouragement design on school choice vouchers (Barnard et al. 2003). This data set is also interesting to us, but is less relevant to our key aim of improving methods for the analysis of drug trials.
These data sets are valuable, but they are dated, and moreover they do not represent the full range of issues that arise in the contemporary statistical study of nonadherence. Thus we would like to use more current datasets to further extend the study of causal inference adjusting for compliance.
We have four specific aims:
1.   To evaluate how changes in the definition of adherence (compliance) affect causal inference results.
2.   To explore how causal inferences that account for compliance vary by strata in the population (e.g., age, smoking status, comorbidities, disease severity).
3.   To devise methods for standardizing causal inferences to match the distributions of stratification factors in the general population, which often differs considerably from the population of clinical trial participants.
4.   To conduct sensitivity analyses to evaluate the extent to which changes in the underlying principal strata (i.e., compliers, always-takers, never-takers) will affect causal conclusions.

Barnard, J., Frangakis, C. E., Hill, J. L., and Rubin, D. B. (2003), “Principal Stratification Approach to Broken Randomized Experiments,” Taylor & Francis, 98, 299-323.
Efron, B., and Feldman, D. (1991), “Compliance as an Explanatory Variable in Clinical Trials,” 86, 9-17.
Goetghebeur, E., and Molenberghs, G. (1996), “Causal Inference in a Placebo-Controlled Clinical Trial With Binary Outcome and Ordered Compliance,” 91, 928-934.
Jin, H., and Rubin, D. B. (2008), “Principal Stratification for Causal Inference With Extended Partial Compliance,” Taylor & Francis, 103, 101-111.

[{ "PostingID": 4720, "Title": "SANOFI-EFC12261", "Description": "A 24-week, Open-label, Randomized, 2-arm Parallel Group, Multinational, Multi-center Clinical Trial to Compare the Efficacy and Safety of Lixisenatide Injected Prior to the Main Meal of the Day Versus Lixisenatide Injected Prior to Breakfast in Type 2 Diabetic Patients Not Adequately Controlled on Metformin" },{ "PostingID": 4729, "Title": "SANOFI-EFC11321", "Description": "Efficacy and Safety of Lixisenatide in Patients With Type 2 Diabetes Mellitus Insufficiently Controlled by Metformin (With or Without Sulfonylurea): a Multicenter, Randomized, Double-blind, Parallel-group, Placebo-controlled Study With 24-week Treatment Period / GetGoal-M-Asia" }]

Statistical Analysis Plan

The publication citation will be added after the research is published.