Palo Alto, CA — A study published this month in JAMA shows how scientists can use our aggregated search engine queries to detect unreported prescription drug side effects and drug-drug interactions before they are found by the Food and Drug Administration’s warning system.
Here’s out the study worked: scientists at Microsoft, Stanford and Columbia University used automated software tools to examine queries by six million Internet users taken from web search logs in 2010. The approach they took is a lot like Google Flu Trends reporting, which uses changes in search behavior to warn of a rise in the prevalence of the flu.
In this case, the scientists weren’t looking for flu symptoms. They were looking for searches connected to an antidepressant, paroxetine, and a cholesterol lowering drug, pravastatin, to determine if taking the two drugs together created any novel side effects.
Their theory: the combination of the two drugs may cause high blood sugar
The scientists were able to explore 82 million individual searches for drug, symptom and condition information. They tracked those searchers over time – first, identifying individual or combination searches for the drugs, then looking for the likelihood that those searchers would also lookup hyperglycemia or ~80 of its symptoms (like “high blood sugar” or “blurry vision.”)
They found that people who had Googled both drug names were significantly more likely to search for terms related to hyperglycemia than were those who searched for just one of the drugs.
By using the way we already manage our own health to learn about the ancillary impact of treatments, these scientists challenged the cumbersome systems we currently rely on.
Think about what we do today: the F.D.A. asks physicians to report side effects through the Adverse Event Reporting System. But it only works when a physician notices a problem and reports it. In clinical trials, patients are often asked about specific symptoms. No surprise that asking people to think about whether or not they’ve had an upset stomach might make them over-anxious and more likely to report the symptom than they would be unaided.
This new approach could supplement those tracking tools with seamless, ambient learning.
Posted by: Leigh Householder
