Pharma R&D Today
Ideas and Insight supporting all stages of Drug Discovery & Development
When drugs collide ….
Posted on July 15th, 2016 by Christy J. Wilson in Pharmacovigilance
Pharmaceutical companies increasingly are adopting “proactive pharmacovigilance” and other strategies to help ensure the safety of a new entity as early as possible in the discovery/development process. But what happens when one safe and effective drug interferes with another because a patient (or clinician) puts the two together, not knowing that a potential drug-drug interaction (DDI) might occur?
By some estimates, DDIs account for more than 30% of all drug adverse reactions. Clinically important events attributable to potential DDI exposure are estimated to affect up to 14.3% of hospitalized patients, and are responsible for up to 0.17% of the nearly 130 million emergency department visits that occur annually in the United States.
In addition, DDIs are a leading contributing cause to drug failure and market withdrawals.
For the sake of patients, the health care system, and a company’s bottom line, prevention is key: it is critical to integrate comprehensive in silico systems that can help detect or predict potential DDIs into the various stages of the drug discovery and development process.
A daunting task
According to recent studies, that’s easier said than done. While no one disputes the utility of in silico modeling for predicting or detecting DDIs, the reliability of such systems depends directly on the comprehensiveness and quality of the data sources used for modeling—and for now at least, there are no universal standards for data input or analysis.
The type and number of potential DDIs also is a challenge. Interactions occur not only among “traditional” drugs, but also with newer entities, such as therapeutic proteins and nanoparticles. And we know that interactions can occur between therapeutic entities and other substances that patients take, such as food, folk and herbal medicines. For example, some research teams are now studying potential DDIs related to Chinese medicine, which in and of itself often includes mixtures of many chemicals with different biological properties.
Not surprisingly, given the scope of the problem, a recent study shows that currently there is no single complete source of potential DDI information that researchers, clinicians or the public can use to direct pharmacovigilance efforts, physician prescribing or self-prescribing. The analysis of publicly available DDI data sources—including clinically-oriented sources, natural language processing corpora and bioinformatics/pharmacovigilance information—demonstrated little overlap across sources. Moreover, the researchers found that comprehensive lists such as RxNorm provided incomplete coverage of potential DDIs that clinicians need to be aware of.
The findings led the study authors to merge data from all of these sources into a single dataset, which they intend to make publicly available for PV text mining. That’s a great beginning.
Revealing hidden DDIs
In another step toward proactively spotting potential DDIs, data scientist Nicholas Tatonetti, a coauthor of the study described above, used text and data mining with his Columbia University colleagues and investigative reporters at the Chicago Tribune as a strategy to identify novel (heretofore undetected or “unseen”) DDIs.
Using sophisticated algorithms to analyze the U.S. FDA’s Adverse Event Reporting System, the team looked for DDIs that might prolong the QT interval—a heart condition that can lead to a potentially fatal arrhythmia. Then they validated their DDI predictions using electrocardiogram data from patient electronic health records (EHRs). The result was the discovery of eight distinct drug pairs that increase the risk of acquired long QT syndrome (LQTS), with different degrees of risk for men and women—and a proof-of-concept that their approach worked.
Other research teams are employing various in silico approaches to identify DDIs early on, including pharmacophore modeling; machine learning techniques; protein-based modeling; and informatics-driven approaches that process input from multiple “big data” sources, including social media. Many in the healthcare industry look forward to the day when their results are standardized and compiled into a universally available resource.
Whether or not a potential DDI becomes an actual DDI depends on many factors, not the least of which are patient-specific risks such as age, genetic variants and polypharmacy. Therefore, clinicians play an important role in evaluating patients before prescribing a drug, monitoring them afterward and documenting the findings in the EHRs.
To maximize the effectiveness of EHRs, clinicians and health systems need to ensure these records are continuously updated and current. As was shown by the LQTS effort, the availability of accurate EHRs—and systems that can mine these records— will enable researchers to identify, validate and reduce DDIs, thereby helping to ensure drug safety.
All opinions shared in this post are the author’s own.
R&D Solutions for Pharma & Life SciencesWe're happy to discuss your needs and show you how Elsevier's Solution can help.
Christy J. Wilson
Sr. Director, Pharma and Biotech Segment
Connect on LinkedIn
Follow on Twitter
- Webinar: Using machine learning to identify adverse events from scientific literature
- Why literature is a valuable source for signal detection: From the EU point of view
- Elsevier uses machine learning to benefit pharmacovigilance
- Data Science Helps Assess the Safety of “Inactive” Ingredients in Drugs
- How Big Data Transforms Reactive Drug Safety to Proactive Pharmacovigilance