Clinical Study Data Request Registered Users, Please Login

Proposal 1528

Title of the Proposed Research

A Swapping Method for Average Treatment Effect Estimation

Lead Researcher

Bimal Sinha


Statistics, the University of Maryland at Baltimore County (UMBC)

Funding Source


Potential Conflicts of Interest


Data Sharing Agreement Date

6 September 2016

Lay Summary

Accurate estimate of treatment effect is vital to clinical studies. A randomized controlled trial (RCT) is the gold standard to reduce bias when testing a new treatment. The people participating in the trial are randomly allocated to either the group receiving the treatment or to a group receiving the control. Such a randomization process minimises the selection bias and therefore allows the researchers to determine any effect differences between the treatment and the control groups. However, the imbalance in covariates such as age, gender, and laboratory measurements can exist in flawed randomized trials due to non-random dropouts and interruption, and non-randomized studies such as open label or single-blind designs. Although methods such as propensity score based methods can remedy the issue, the imbalance continues to be one of major issues in clinical studies.

We propose a swapping method to achieve the balance. We classify all patients into subclasses based on discrete covariates such as gender. Within each subclass, we then estimate the average treatment effect based on the principle of choosing a better model between the treatment and baseline groups. The simulation studies indicate that our method provides more stable estimates than traditional methods. The proposed datasets are for non-randomized studies. They are helpful to illustrate the public the importance of correcting the bias and therefore providing more reliable and accurate estimate of treatment effects in clinical studies. We can provide a powerful tool that will be published and distributed to the community.

Study Data Provided

Study GSK-GLP108486: A Randomized, Open-Label, Active-Controlled, Parallel-Group, Multicenter Study to Determine the Safety and Efficacy of Albiglutide Administered in Combination With Insulin Glargine as Compared with the Combination of Insulin Glargine and Preprandial Lispro Insulin in Subjects With Type 2 Diabetes Mellitus
Study LILLY-H9X-MC-GBDE: A Randomized, Open-Label, Parallel-Arm Study Comparing the Effect of Once-Weekly Dulaglutide with Once-Daily Liraglutide in Patients with Type 2 Diabetes (AWARD 6: Assessment of Weekly AdministRation of LY2189265 in Diabetes-6)
Study ROCHE-BO16216: A randomized, open-label study of the effect of Herceptin plus Arimidex compared with Arimidex alone on progression-free survival in patients with HER2-positive and hormone-receptor positive metastatic breast cancer
Study ROCHE-BO21977: A randomized, Multicenter, Phase III Open-Label study of the efficacy and safety of Trastuzumab-MCC-DM1 vs. Capecitabine + Lapatinib in patients with HER2-Positive locally advanced or metastic breast cancer who have received prior Trastuzumab-based therapy.

Statistical Analysis Plan

Publication Citation

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

Summary Results

Results summary or link will be posted when available.