The Food and Drug Administration is talking to Google about how the search engine could help the agency identify previously unknown side effects of medications. Agency officials held a conference call on June 9 with a senior Google researcher who co-wrote a 2013 paper about using search query data to identify adverse drug reactions, according to a record of the meeting posted to the FDA website that hasn't been previously reported. Microsoft researchers also say they have been working informally with the agency for several years on detecting drug side effects.
FDA spokesman Chris Kelly called the meeting an introduction and a chance "for the agency to begin a discussion on how we might collaborate with Google on identifying adverse event data, using Google’s technologies and data.” The agency declined to make officials available for interviews, and Kelly wouldn't comment on the FDA's talks with other companies. A Google spokesman had no comment.
The Google scientist on the call was Evgeniy Gabrilovich. His bio on Google’s research page says he’s a senior staff research scientist specializing in data mining. A former employee of Yahoo, Gabrilovich co-wrote a paper two years ago that used Yahoo search data to identify suspected drug reactions. The analysis, based on 176 million Yahoo queries in 2010, demonstrated that search data can help find drug reactions "that have so far eluded discovery by the existing mechanisms," according to the paper. It was published in the peer-reviewed Journal of Medical Internet Research.
Before drugs are approved by the FDA, the only people who get them are carefully selected patients enrolled in clinical trials—generally a few thousand people at most. After they reach the market, medicines may go to hundreds of thousands, or even millions of people. Some of them will be taking additional pills or have conditions that the drug affects. Evidence of negative side effects can lead regulators to change a drug’s safety warnings or prescribing practices. In rare cases, safety concerns can cause a a medicine to be pulled from the market entirely, as happened with the painkiller, Vioxx.
The government’s process for tracking so-called adverse events (which involves patients, doctors and pharmaceutical companies submitting forms that describe possible reactions) hasn’t changed much since the late 1990s. The FDA now gets more than a million reports of adverse drug reactions a year. Although the agency has tried to make the data easier to access, critics say the system probably misses many adverse events and can be slow to detect safety problems.
Companies have been trying for years to sift the noise of the internet for meaningful signs of drug reactions. “If you have the right technology to connect the dots, then you can see problems very, very early on,” says Ido Hadari, chief executive of Treato, which scans patient forums and other online postings. He wouldn’t comment on whether Treato is talking with the FDA.
Beyond the discussion with Google, there are signs that the FDA is expanding its search for new sources of information and ways to monitor the safety of drugs on the market. Last month the agency announced a collaboration with PatientsLikeMe, an online patient network. And last year, an FDA researcher co-authored a paper about monitoring drug safety on Twitter.
Microsoft's researchers have been working on the problem for several years and have co-authored a paper with FDA colleagues, says Eric Horvitz, distinguished scientist and managing director at Microsoft's research arm.
Horvitz, along with other researchers from Microsoft and Stanford University, published a 2013 paper (PDF) finding that Web search data could have exposed the adverse interaction between the antidepressant Paroxetine (aka Paxil) and cholesterol-lowering drug Pravastatin, which together can cause hyperglycemia, or high blood sugar. People who searched for both of those drugs over a 12-month period were also more likely to search for terms related to high blood sugar, such as diabetes and dry mouth, according to the paper in the peer-reviewed Journal of the American Medical Informatics Association. The analysis of millions of searches1 on Google, Yahoo, and Microsoft's Bing were from 2010, before the interaction was publicly reported the following year.
Google scientist Gabrilovich’s paper analyzed how searches2 for such common symptoms as “cramps,” “weight gain,” or “tired” differed among people who also searched for the name of a medication. While serious symptoms that appeared shortly after treatment started were likely to be known side effects reported to the FDA, the researchers found that search data were more likely to reveal reactions that "appear much later after the beginning of treatment, hence their association to the drug is often overlooked."