Advancing methods for applying the results of clinical trials to data from medical insurance claims for patients taking anti-coagulants for non-valvular atrial fibrillation

Atrial fibrillation is a condition where the heart beats irregularly and often too fast. A common form of this condition, called non-valvular atrial fibrillation (NVAF), affects millions of people around the world. NVAF increases the risk of dangerous blood clots forming in the heart. These clots can travel to the brain and cause a stroke by blocking a blood vessel, or to other parts of the body, leading to serious complications.

To lower this risk, patients are usually prescribed blood-thinning medications known as anticoagulants. Warfarin was the traditional drug used for many years, it works by blocking the action of vitamin K which is required by the liver to produce certain blood clotting factors. However, newer medications—called direct oral anticoagulants (DOACs)—are now more commonly prescribed. DOACs work by directly targeting and inhibiting specific proteins involved in the clotting process. Two examples of these are dabigatran and edoxaban, which have been shown in clinical trials to be as effective as warfarin for preventing strokes in patients with NVAF.

However, clinical trials often involve very specific groups of people, which means the results may not always apply to everyone who takes the drug in the real world. Our research will focus on understanding how well the findings from two large clinical trials—RE-LY (which studied dabigatran) and ENGAGE-AF-TIMI 48 (which studied edoxaban)—can be applied to a broader group of people who were not part of those trials.

To do this, we will use a large database of U.S. medical insurance claims called the IBM MarketScan Research Databases. This database contains de-identified health information from millions of patients and will allow us to create a large group of people who would have been eligible to take either DOACs or warfarin. We will assess how the outcomes from this real-world group compare to those in the trial data and use advanced statistical techniques to simulate the trial results in the real-world group.

This work is important because it will help improve how results from clinical trials are used to guide treatment decisions for the wider population. It will also explore better ways to use real-world data to support future trials and patient care. Importantly, our methods use synthetic, computer-generated data and do not require linking personal information, so there is no risk of identifying individuals.

We will develop and test these methods to improve the accuracy and usefulness of trial results for patients who were not part of the original studies. This research has the potential to make clinical trial findings more relevant and helpful in everyday medical practice.