Researchers at Columbia University are leveraging machine learning to identify and predict differences in adverse drug effects between men and women. Nicholas Tatonetti, an Associate Professor in the Department of Biomedical Informatics and computer science undergraduate Payal Chandak had developed an algorithm called AwareDX (i.e., Analyzing Women At Risk for Experiencing Drug toxicity), by examining 50 years’ worth of reports in the US Food and Drug Administration (FDA) database. Related findings were published in Patterns recently.

A proactive approach to tackle a lack of heterogeneity

According to the two researchers, adverse drug reactions are the fourth leading cause of death in the US. While women usually take a longer time to metabolize medications and twice more likely at risk of developing an adverse drug reaction as compared to men, the actual differences had not been thoroughly understood or quantified.

Real world clinical data has provided an opportunity to estimate safety effects in these otherwise understudied populations but there has been a scarcity of such information because all along, clinical research trials were done on a rather homogenous population (i.e., healthy White males). For example, Ambien, a sleep-aid which was previously prescribed at the same dosage for both men and women. After repeated evidence claiming women could experience a significantly more adverse reactions the following morning, FDA recommended to cut the dosage by half in 2013.

“Rather than take the stance that we wait for evidence to become so overwhelming that we have to do something about, we wanted to be more proactive,” Tatonetti says. “We want to use databases like the Adverse Event Reporting System (FAERS) from the FDA or the electronic health records (EHRs) to get a jump on identifying sex-specific adverse events before it’s too late”.

An algorithm which mitigates gender biases and affects health

FAERS contains reports of adverse drug effects from consumers, healthcare providers and manufacturers dated back to 1968 and what AwareDX did was to digest and group the data into sex-balanced subsets before looking for any pattern or trend. The algorithm repeated the process for 25 times to improve results.

As some drugs or effects are more commonly use or seen in either sex, the algorithm also addresses and mitigates these confounding biases by having a statistical basis to identify sex differences in adverse drug reaction and rank them by strength of the evidence. For example, the algorithm has successfully validated the ABCB1 gene can pose different risks to men and women while they are taking different types of medications. The gene will result in a higher risk of slow heart rate for women taking risperidone, a kind of antipsychotic drug and a higher risk of muscle aches for men taking simvastatin, a cholesterol medication.

Overall, researchers were able to bank more than 20,000 of such potential sex-specific drug effects. Researchers believe as the algorithm continues to improve, it may assist doctors to look at the adverse effect information specific to the drug they prescribe. Once the information is further studied and validated, it will actually impact drug prescriptions and people’s health.


Author Bio

Hazel Tang A science writer with data background and an interest in the current affair, culture, and arts; a no-med from an (almost) all-med family. Follow on Twitter.