For doctors or patients who notice side effects to a prescription drug (or a combination of several), there’s one place to go: the FDA’s Adverse Events Reporting System. This is where, once a drug is already on the market, the government can monitor side effects that for one reason or another did not turn up in trials.
The problem is that many patients aren’t so forthcoming, and may not report — to their doctors or the FDA — minor side effects they notice.
But they will tell someone something: Google (or Bing or Yahoo).
According to a new study in the Journal of the American Medical Informatics Association, more than one in 250 people (0.43 %) searched for one of the 100 best-selling drugs at some point in 2010. And if a whole bunch of those people searching for two of those drugs also search for, say, blurry vision, well, you might have just found out that that particular combination of drugs can cause hyperglycemia.
And that’s just what researchers from Microsoft, Columbia, and Stanford found. Digging into a huge mountain of data harnessed from the search bars of consenting users, they found that people who searched for *both* paroxetine (a common antidepressant) and pravastatin (for those with high cholesterol) were more likely to also look for a list of some 80 terms describing symptoms associated with hyperglycemia (pdf) than those taking just one of the medications, who in turn were more likely than those taking neither.
One question the paper examines is whether it’s possible that the results may be skewed by anxious people who Google every which thing they worry about — and paroxetine is commonly prescribed for anxiety. But, they write, “the data do not support this potential explanation … [as] there is less of an effect for those who search for paroxetine alone.”
The authors liken the use of web-search data to “a large-scale distributed network of sensors for identifying the potential side effects of trugs.” Search data won’t replace other traditional systems of side-effect monitoring, but it can complement our existing systems. After all, searches are about as direct a line to patients as you’re going to get — and researchers are ready to listen, or, perhaps more accurately, compute.