In what might possibly be seen as a self-fulfilling prophecy, researchers have used internet search data to determine side effects for drugs not otherwise known when the drugs were initially developed. Such a search would be independent of the reporting mechanism the Food and Drug Administration (FDA) uses to identify side effects post-testing. The specific focus of the study ($), which was a collaboration between researchers at Stanford, Microsoft and Columbia, looked at search queries to determine possible side effects of combinations of drugs. This follows on similar work looking through the FDA’s Adverse Event Reporting System for drug-drug side effects. The study indicated side effects could be found in an automated way, which is a bit more thorough and perhaps reliable that the FDA system, which is voluntary and relies on physician input, rather than feedback directly from the patients.
The parallel here is Google Flu Trends. However, search engine data as a predictor of health concerns is inexact. Google Flu Trends did a horrible job estimating this year’s outbreak, possibly because news coverage of flu in January skewed the analysis of subsequent search data.
It certainly is possible that some of the side effects found via search data will not present themselves at the anticipated rate due to searches prompted by reasons other than the searcher experiencing them directly. But since this work is focused on drug-drug interactions, it’s possible that the smaller affected populations (those using both drugs) may control for the kind of discrepancy Google Flu Trends found this year. More analysis needs to be done, and I think it fair to say that the FDA (or in the case of Google Flu Trends, the Centers for Disease Control) shouldn’t forego their own reporting systems in favor of mining online search data. Having an additional tool is great, but there’s not yet a reason to throw out what’s worked in favor of what might be faster, if not as reliable.