Clinical Study Data Request
Please note that this site will be unavailable from Friday 23 February to Monday 5 March 2018 for a major upgrade, in order to improve the site for researchers such as yourself who request
patient-level data.
Registered Users, Please Login

Proposal 1295

Title of the Proposed Research

A Swapping Method Based on Covariate Classification for Average Treatment Effect Estimation.

Lead Researcher

Dr. Bimal Sinha


University of Maryland, Baltimore County (UMBC), Department of Mathematics and statistics

Funding Source

Rowena is supported by Teaching Assistantship, and the other two major researchers are self-funded.

Potential Conflicts of Interest

The lead Researcher declares that he has no competing interests. Researcher 1 declares that she has no competing interests. Researcher 2 declares that she has no competing interests. The Statistician declares that he has no competing interests.

Data Sharing Agreement Date

2 October 2015

Lay Summary

A swapping method is a novel method to improve the estimate of average treatment effect in clinical trials and observational studies. The average treatment effect is the estimated treatment effect compared to the baseline. However, the imbalance in covariates, the measures other than the treatment/baseline and the patient response, exists in non-double-blind randomized trials, and can bring in the bias in estimating the average treatment effect. Although traditional methods can remedy to the issue, the imbalance continues to be an issue.

We propose a swapping method to achieve the balance via two steps. We first classify all patients into subclasses based on discrete covariates. Within each subclass, we then calculate the average treatment effect using the better model between the treatment and baseline groups. The simulation studies show that the proposed method provides more stable estimates than traditional methods.

The applied datasets are important to help the public understand that the imbalance in covariates exists in non-double-blind randomized trials. Therefore, the powerful tools are needed to correct the bias in better estimating the average treatment effect. The effective estimate of the average treatment effect is important in making the decision on the treatment selection and the clinical trials. We provide powerful solutions and tools for estimating the average treatment effect in clinical trials and observational studies.

Study Data Provided

Study BI-1235.7: Open Label Study Telmisartan and Amlodipine in Hypertension
Study BI-1200.10: An Open Label Phase II Trial of BIBW 2992 in Patients With HER2-negative Metastatic Breast Cancer
Study GSK-NPP30006: Open label, safety study for use of lamictal in patients with diabetic neuropathy

Statistical Analysis Plan

The statistical analysis plan will be added after the research is published.

Publication Citation

A manuscript will be submitted for publication in 3 months, and the results can be upload to if required once the paper is accepted.