Fragility Index (FI)

What is it?

The fragility index is a measure of the “fragility” of a clinical trial. Using a 2x2 table using a fisher’s exact test as an example, it is equivalent to the required number of people who, if we switch their outcome in the smaller group, would cross the critical value for the statistical test used to result in a non-significant result.

Project’s Motivation

The related work section has links to different publications that sparked this project. Briefly, the Fragility Index has been suggested to be an easy-to-understand metric of a clinical trials’ robustness, which may pair well with other frequently discussed metrics like p-values. There have also been critics to its use or potential misuse. We are interested how the fragility index may differ based on different clinical trial characteristics. Multiple papers have shown that it requires just a few patients to alter outcomes of many large phase III trials. It is of concern to us that these studies require billions of dollars sunk but outcomes can be “easily” rendered obsolete based on so few participant outcomes.

Initial Hypotheses

Our main question of interest is whether fragility index is associated with a disease type. Specifically, we want to test the difference in fragility index for cancer related therapies to non-oncology immunologic monoclonal antibodies such as allergy, diabetes, and asthma. As a prior belief, we hypothesize that the median fragility Index of trials targeting cancers will be less than trials targeting auto-immune chronic diseases.

We hypothesize that there will be a difference between the median fragility index of trials targeting cancers and trials targeting auto-immune chronic diseases. We believe the fragility index of non-oncology trials will be much higher. We believe this to be true for two main reasons:

• There is more at risk in treating cancer than most autoimmune disease cases. The outcome in cancer is death while autoimmune diseases tend to cause chronic illnesses and/or decreased quality of life. We believe that this change in outcome will be seen in riskier trials existing in cancer compared to autoimmune diseases.
• Cancer is thought anecdotally to be ‘ahead of the field’ in its research/treatments and has the luxury of additional funding. We believe being further on the cutting edge will result in more fragile trials.

We have a few secondary questions of interest if time allows:

• If a drug is already approved for one indication, we hypothesize that the fragility index of the next trial will be much higher. We believe this for two reasons:
  • A drug company would not make the 2nd trial in an already existing drug if they did not have full confidence it would work.
  • In between a drug being approved for one indication and the next, doctors who are affiliated with research may prescribe therapies off label. There may be a host of evidence in the field from doctors using the drug off label, which gives data to suggest the drug may work in a different indication before it is approved.
• Is a drug intended to treat children less fragile than other drugs?

What did we do in this project?

Due to realities explained in the Obtaining Data section, we did not answer any of the questions initially explored above. The data we collected was insufficient to answer these questions.

Instead, this project:

1. Explored various methods of data scrapping for clinical trial data
2. Calculated the fragility index for data we were able to collect
3. Created an interactive FI calculator
4. Created a graphical display to show geographical and temporal trends in phase 3 clinical trials

 

---

A work by Bryan Bunning, Yuanzhi Yu, Zongchao Liu, Gavin Ko, and Kevin S.W.

© Copyright 2019 The Fragility Index Project Team, All Rights Reserved | Powered by Github.io