20 September 2023

Giant study finds 694 new potentially harmful drug interactions

Drug interactions

Data analysis of more than one million Danish inpatients identifies new drug interactions that are associated with mortality, hospital readmission or longer hospital stays.

In a new study, professor Søren Brunak and his colleagues identifies 4,000 drug pairs that are associated with more frequent dosage adjustments when prescribed together. Photo: Canva.

In some cases, especially among elderly populations, a patient may be prescribed several different medications at once to treat one or multiple health conditions. This phenomenon is known as polypharmacy.

Due to possible harmful interactions between different drugs, polypharmacy is associated with health risks. Research into polypharmacy has typically focused on the number and type of drugs a patient receives due to lack of access to millions of patient records.

However, the dosage of each drug also influence health outcomes.

In a new study, professor Søren Brunak and his colleagues identifies 4,000 drug pairs that are associated with more frequent dosage adjustments when prescribed together.

“These pairs were also associated with higher levels of hospital readmission, mortality, and longer hospital stays, and 308 were associated with worsened kidney function,” says Søren Brunak.

“I want to underline that this study does not prove any cause-effect relationships, but it suggests that some of these drug pairs may involve previously undiscovered interactions. The drugs we identified are often given to fewer patients and therefore harder to detect in small scale studies – contrary to the 185 million treatment episodes used in this study. Future research will be needed to explore that possibility,” says Søren Brunak.

Analyzed data from more than one million patients

To examine polypharmacy in the context of drug dosages Søren Brunak and colleagues analyzed data from the electronic health records of more than one million patients admitted as inpatients to Danish hospitals from 2008 to 2016.

They used a statistical approach known as Bayesian inference to identify drug pairings associated with more frequent adjustments to the dosage of the drugs.

“Out of 77,249 total drug pairings initially tested, 3,993 were associated with more frequent dosage adjustments when given together,” says Cristina Leal, the first author of the paper.

Previous research has already linked many of the drug pairs associated with dosage adjustments to potential harmful drug interactions. These may be seen as nice positive controls. However, of the significant pairs, 694 may represent potential new, undiscovered drug interactions.

 “The next step will be to characterize specific adverse events in the patient records by text mining, and feed that information back to clinicians and the Danish Medicines Agency when relevant,”  says Cristina Leal.

Søren Brunak pinpoints that the study is an example of how one can make discoveries with Danish healthcare data due to their high quality and availability.

“Moreover, we can keep the data in highly secure supercomputer environments that make it possible to finish large-scale analyses in a reasonable amount of time,” he says.

You can read all of “Dosage tweaks may hint at undiscovered interactions between medications” in PLOS Digital Health.


Professor Søren Brunak

Journalist and press consultant Liva Polack
E-mail: liva.polack@sund.ku.dk