FDA proposes data mining as one technique useful for drawing connections between products and adverse events in a draft guidance on post-marketing safety surveillance.

The ¹ post-marketing guidance states that data mining, or "looking systematically into the data by using statistical or mathematical tools," may enable sponsors to "identify unusual or unexpected product-event combinations warranting further investigations."

The draft guidance, entitled "Good Pharmacovigilance Practices and Pharmacoepidemiologic Assessment," is based on a similar concept paper released in March 2003, although the data mining section is new (² Also see "Pharmacoepidemiologic Studies Should Estimate AE Background Rates – FDA" - Pink Sheet, 10 March, 2003., p. 4).

The section was added because "this is very important to us and also to the industry," Center for Drug Evaluation & Research Acting Director Steven Galson, MD, said in a May 4 conference call announcing the release of the draft.

Data mining is "happening a lot, and we wanted to emphasize how it can help identify product-adverse event combinations in our large database and how best to interpret the data."

A working group created by FDA and the Pharmaceutical Research & Manufacturers of America plans to release a "white paper" outlining best practices for data mining (³ Also see "FDA Moves Ahead With Adverse Event Data-Mining; “White Paper” Out Soon" - Pink Sheet, 26 January, 2004., p. 11).

The Center for Biologics Evaluation & Research has initiated data mining projects for biologics and vaccines, according to the center's fiscal year 2003 annual report.

CBER is "partnering with health care systems and others to use external medical databases that provide rapid and reliable information on product safety," the report states. A pilot project assessing safety of vaccines given to older adults is underway. The center is also developing new data mining tools through a partnership with an "expert constituent group."

Because large adverse event databases such as FDA's Adverse Event Reporting System are subject to a number of biases, results of such data analysis should be considered "hypothesis-generating," the draft guidance states.

"Further investigation of a product-event combination may be warranted, especially if the event is serious and unlabeled or raises other safety concerns."

Such safety signals include an apparent increase in the severity of a labeled event, more than a small number of serious events thought to be extremely rare, or identification of a previously unrecognized at-risk population.

"Signals warranting additional investigation can be further evaluated through carefully designed observational studies of the product's use in the 'real world,'" such as pharmacoepidemiologic safety studies, registries and surveys.

FDA also recommends sponsors put the signal events in the context of a national denominator of patients exposed to the product.

Although there are limitations to such a calculation, "we recommend that sponsors calculate crude adverse event reporting rates as a valuable step in the investigation and assessment of adverse events," the draft guidance states.

FDA issued two other draft guidances May 4. A ⁴ guidance on developing risk minimization plans loosens a previously proposed system for risk management plans (see ⁵ [A#00660190011] ).

A ⁶ guidance on premarket risk assessment is consistent with a 2003 concept paper (⁷ Also see "Rx Name Risk Assessments Should Be Conducted By Manufacturers, FDA Says" - Pink Sheet, 10 March, 2003., p. 5).

The three risk management guidances were mandated by the user fee program reauthorization in 2002. FDA aims to issue final guidances by the legislative deadline of Sept. 30. The 60-day period for comments on the drafts closes July 4.

The agency said the guidances are not intended to establish new regulatory burdens.

"What we're really doing is laying out in a more transparent way what we're already doing. This isn't imposing any new requirements," Galson said.

Post-Marketing Safety Surveillance Should Include Data Mining, FDA Says

More from Archive

More from Pink Sheet