Carnegie Mellon University

drug tablets

The Benefit of Using Data Analytics to Select Drug-Disease Pairs for Clinical Trials

By Sridhar R. Tayur, Ford Distinguished Research Chair; University Professor of Operations Management

In 2017, the Food and Drug Administration announced a fast approval process for new uses of drugs already in the market—if it can be shown through existing data (and a small clinical trial, if needed) to be effective for a different disease.

This is a boon to drug companies. Clinical drug trials are very expensive to run, and avoiding a full-blown trial saves money and time—while saving more lives earlier and boosting profits.

Data Analytics and Machine Learning Improve Trial Efficiency

Companies are looking for the highest probability of success with the least amount of additional work. One way to do this is to extract the most information they can from the enormous amount of data they already have or can readily obtain. This is called observational data.

It is an untapped gold mine. Much of it, however, is confounded data. This is where our work in data analytics and machine learning comes into play.

Using a well-known data analytics approach known as causal inference, we develop an efficient procedure to select a drug-disease pair to perform clinical trials. Until now, causal inference research did not focus on scenarios where large amounts of confounded data could later be selectively deconfounded (via an observational study). We’re the first to show that by smartly incorporating the confounded observational data, you can actually reduce the number of samples needed to estimate the effectiveness of the drug in treating the disease.

Improved Testing Efficiency Leads to Higher Chance of Success and FDA Approval

Our paper develops the theory and offers three different ways to select the patients. We tested our approach on 96 different choices of cancer and patient characteristics, such as age. We find that by carefully selecting patients who have certain characteristics, we can reduce the number of patients we need for estimating the treatment effect accurately.

For companies looking to quickly reposition their products, our enhanced causal inference method unlocks the potential of observational data in the most efficient manner—leading to drug-disease pairs with a higher chance of success and FDA approval.