broad perspective to a fine focus, this ability for the
user to ‘drill down’ to the level of detail in which
he/she is interested can be helpful. Furthermore, a
graphical display of the data can help to identify
data problems, provide insights not achievable
with mere tables and demonstrate new relation-
ships.
40.9 Summary
Data mining does not supplant traditional pharma-
covigilance methods. Instead, it supplements safety
surveillance methods and allows a systematic iden-
tification of potential safety signals. The promise of
data mining using large regulatory safety databases
is that the huge size and diversity are the primary
advantages because they enable multiple compari-
sons and provide valuable information whether one
is looking in groups of events or classes of drugs.
Essentially it allows the review of all of the data and
is very sensitive in detecting safety signals.
The interpretations of data mining results need
the expertise of safety reviews and medical officers
to analyze and interpret data appropriately. Data-
mining signals by themselves are not indicators of
problems, but are indicators of possible problems.
Moreover, caution must be exercised with any
comparison of disproportionality ratios across dif-
ferent products, for example comparison with
competitor drugs because of thevarious limitations
thatexistwhen making these kinds of comparisons.
Finally, all signals should be evaluated recogniz-
ing the possibility offalse positives. In addition, the
absence of a signal does not mean that a problem
does not exist.References
Bate A,et al.1998. A Bayesian neural network method
for adverse drug reaction signal generation.Eur. J.
Clin. Pharmacol. 54 : 315–321.
DuMouchel W. 1999. Bayesian data mining in large
frequency tables, with an application to the FDA
spontaneous reporting system.Am. Stat. 53 : 177–
190.
Evans SJW,et al.2001. Use of proportional reporting
ratios (PRRs) for signal generation from sponta-
neous adverse drug reaction reports.Pharmacoepi-
demiol. Drug Safety 10 : 483–486.
FDA Guidance for Industry. 2005.Good Pharmacov-
igilance Practices and Pharmacoepidemiologic
Assessment (http://www.fda.gov/cder/guidance/
6359OCC.pdf).
FDA Guidance for Industry. 2005. E2E Pharma-
covigilance Planning. Step 4 of the ICH Process,
November 2004(http://www.fda.gov/cder/guidance/
6355 fnl.pdf).
Frawley W,et al.1992. Knowledge discovery in data-
bases: an overview.AI MagazineFall: 213–228.
HHS OCR HIPAA Privacy. 2003.General Overview of
Standards for Privacy of Individually Identifiable
Health Information(http://www.hhs.gov/ocr/hipaa/
guidelines/overview.pdf).
Szarfman A,et al.2002. Use of screening algorithms
and computer systems to efficiently signal higher-
than-expected combinations of drugs and events in
the US FDA’s spontaneous reports database.Drug
Safety 25 (6): 381–392.REFERENCES 555