About

Background

Adverse Drug Effects (ADEs) are unwanted side effects from taking drugs. These can range from being minor effects (eg. drowsiness) to having very serious and life-threatening consequences. Many occurrences of ADEs are due to the inherent nature of the chemical structures of drugs; single drugs can have unwanted side effects, but this is pretty well researched and documented.

However, multiple drugs taken together can interact with each other to cause further ADEs. These drug-drug interactions (DDIs) are not well researched or documented, though they are prevalent and have a huge cost on the healthcare system. For example, DDIs have accounted for ~26% of ADEs in the US, have occured in half of all hospital patients, and cause 74000 ER visits and 195000 hospitalizations every year (Chasioti et. al, 2018). Though DDIs are a major culprit in ADEs, we recognize that it is not possible to distinguish between an individual drug and a DDI as the cause of an ADE for a variety of reasons.

Purpose of this Website

This website is a framework for researchers to use to investigate DDIs and their effects on the ADE myopathy. We have taken raw patient data and produced statistics and visualizations from which researchers and medical practitioners (and potentially patients) can draw insights. To accomplish this, we have focused on two logical components - 1) formatting aggregated datasets and applying data analysis and computational methods, and 2) presenting the results of this analytical process in accessible and compelling visualizations, alongside other relevant information and findings. Users can interact with this framework using a public web application. In the future, we intend to allow users to upload or select different datasets for examination and be able to run the analysis and query for visualizations and results.

In this application, have used data from the FDA Adverse Effects Reporting System Database (FAERS Database) as a test dataset. The FDA publishes the data we are using under the condition that it should not be used for clinical decision making. As we are unable to obtain EHR data, we are developing this project as a functional framework, which will allow users with trustworthy (e.g. less statistically biased) data to use the application to ultimately gain insights which could be used in clinical decisions.

What can I do on this Site?

• Look up odds ratios between drug subsets from a drug group of up to size 6.
• Find the active ingredients in a common (generic) drug.
• Find additional drug information such as brand name, packaging, listing expiration date, and dosage form.
• Find side effects of drugs.
• Find a list of all drugs in our database (see FAQ #7).

FAQ

1. What does the term Odds Ratio mean?

The Odds Ratio is a statistic that indicates the increased chances of a side effect in the presence of some variables. In the context of drug combinations, it measures the increased chances of an ADE when adding some set of drugs to a patient’s current prescription sets. For example, an OR of 2.5 for a single drug means that the odds of a patient experiencing the adverse effect when taking the drug are 2.5x as high as the odds of the patient experiencing the adverse effect without taking the drug. An OR of 2.5 from a drug set Set A to a superset Set B means that the odds of a patient experiencing the adverse effect when taking all of the drugs in Set B are 2.5x as high as the odds of the patient experiencing the adverse effect while taking all of the drugs in Set A - the drugs in Set B that are not in Set A are the added drugs associated with increased risk.

2. What is the importance of myopathy in this context?

In the initial version of our website, we have chosen to focus on drug combinations associated with a patient's increased risk of experiencing a specific ADE - myopathy. All combinations give Odds Ratios with respect to a patient's increased risk of myopathy.

3. How do I interpret the Sunburst visualization?

The Sunburst plot is a hierarchical visualization depicting the Odds Ratio between every subset of drugs of your input drug set. Each piece of a ring on some layer L represents a subset S of drugs. In the layer L' that immediately surrounds L, any other piece S' whose area is within the boundary of piece S is a subset of S. Hovering over piece S' shows the Odds Ratio going from S to S'. The color of the ring indicates the value of the Odds Ratio. You can see the value from the legend below the plot.

4. How do I search up a drug combination?

First click Find Odds Ratio. A popup window will appear where you can search for drugs to add to your working set of drugs. After adding the drugs that you want, press Compute OR.

5. How can I remove a drug from my working set of drugs?

Click the X button next to the drug to remove it from your set.

6. Can I find Odds Ratios between more than 6 drugs??

Our current implementation only computes Odds Ratios between 1 and 6 drugs.

7. Can I use common (generic) names in my odds ratio search?

Our current method looks only for scientific names. If you don't know the scientific name, try using the Find Active Ingredients feature to in the Combination popup window to find scientific names of active ingredients. If you have entered an alias to a drug in our database, the 'Compute OR' operation will also attempt to convert this (behind-the-scenes) to one of our standardized drug names before running the calculation.

8. Is there a list of drugs that are in your database?

Yes. Navigate, using the top navigation bar, to the Drugs List page.

9. What browsers are supported for viewing this application?

We have tested this application in Chrome and Safari. If you are having difficulty viewing the application, please try accessing it in one of those browsers.

Disclaimer

Our tool uses FAERS data, which should not be used for clinical decision making.

Acknowledgements

Built in 2018-2019 by William Glisson, Yuan Kong, Andrew Wang, and Andy Ye, computer science students at the University of Pennsylvania for their senior design project. Many thanks to Professor Li Shen and Xiaohui Yao from the Shen Lab at the Perelman School of Medicine at the University of Pennsylvania for their mentorship, as well as Professor Ani Nenkova from the Computer and Information Sciences department at the University of Pennsylvania.