Data mining for signals in spontaneous reporting databases: proceed with caution.

Purpose: To provide commentary and points of caution to consider before incorporating data mining as a routine component of any Pharmacovigilance program, and to stimulate further research aimed at better defining the predictive value of these new tools as well as their incremental value as an adjunct to traditional methods of post-marketing surveillance.

Methods/results: Commentary includes review of current data mining methodologies employed and their limitations, caveats to consider in the use of spontaneous reporting databases and caution against over-confidence in the results of data mining.

Conclusions: Future research should focus on more clearly delineating the limitations of the various quantitative approaches as well as the incremental value that they bring to traditional methods of pharmacovigilance.