Stanford algorithm uncovers drug side effects

When doctors believe a drug's side effects are behind a patient's aches and pains, they can report their suspicion to the FDA. But it can be hard to know whether those adverse symptoms are really drug-related or the result of another illness. Now Stanford researchers have developed a computer algorithm to show which adverse events are likely to be drug-related, shining a light on new bad reactions to meds as well.

They let their algorithm loose on the FDA's large database of patient and physician adverse-event reports and made multiple discoveries. A key finding was that a class of popular antidepressants called SSRIs (think Lexapro and Prozac), when taken along with blood-pressure pills known as thiazides, are associated with an effect that is linked to irregular heart beats and sudden cardiac death, according to Stanford.

Based on their findings, the Stanford researchers have set up two databases of their own that are available to the public: one that includes hundreds of new side effects for each of the more than 1,000 drugs covered in the database, and the second that exposes new bad interactions between drugs.

Why is this important? Elderly and chronically ill patients often take a battery of meds to keep their health in check, and, obviously, sometimes doctors prescribe drugs that might not get along with others in their patients' systems without knowing the potential side effects from the combos. These are the types of insights into the real-world effects of drugs that can be missed in clinical trials, too.

"It sounds obvious, but it's a nifty statistical way to eliminate bias," said Nicholas Tatonetti, a grad student at Stanford, who was lead author on the group's paper on the research in Science Translational Medicine. "And we found that the more things you can match between the groups, like other drugs the people have in common, the more likely you are to also unintentionally match for variables you may not have even thought about but that may affect the result."

- here's the release
- see the piece from a Stanford blog