Searches on 3 prominent search engines allowed researchers to pinpoint unreported drug side effects prior to official notices from the FDA.
The study, conducted by Microsoft, Stanford University, and Columbia University researchers, was published online March 6, 2013, in the Journal of the Medical Informatics Association
. The New York Times
also included a summary of the report on that date.
According to the study abstract, the authors aimed to determine whether Internet searches could provide early clues about adverse drug events, and studied web search data logs to find queries on paroxetine and pravastatin interactions.
Researchers were able to find a hyperglycemic reaction prior to the FDA’s official Adverse Event Reporting System alert, they said. The search engine logs are also inexpensive to collect, adding to their potential benefit to drug safety surveillance, the abstract added.
The New York Times
noted that the study used data mining techniques similar to those used by Google Flu Trends, the search engine–created flu prevalence tracker. It refines the work of researcher Russ B. Altman, of Stanford University’s bioengineering department, in automating the hunt for drug–drug interactions within the data in FDA reports.
Currently, the Adverse Event Reporting System generates its data only when a physician notices and reports a reaction, The Times
In the latest study, researchers isolated searches for each specific drug, and searches for both drugs, and then determined the likelihood that users would search for hyperglycemia and its symptoms.
The signals were particularly strong, with those who searched for both drugs being more likely to search for hyperglycemia and its related terms.
The results have researchers considering new sources of information, including information from social media and behavioral data, The New York Times
stated. The challenge, however, is guarding personal privacy while incorporating new data sources, it added.